Hey agdr,
did you calculate the values for the capacitors and if yes for what voltage drop?
A few of the bypass cap choices seem a bit strange, mostly the epcos and os-con, not sure how I feel about tantalum.
Any reason for the 10 4.7µF film decoupling caps? Pricing in the US seems inflated as it is, why not get fewer but bigger caps instead? Concerned about high frequency AC rating on the 6.8µF ones? Pricing/Availability? Why not MKS4? Or simply because you started with them and then added more and more?
And a not capacitor related question: Are you aware of the NJM4556's DC offset drift with temperature and just don't consider it an issue with the power dissipation of the SIP8 package?
Don't get me wrong, I'm just trying to understand the design choices, not trying to talk it down.
did you calculate the values for the capacitors and if yes for what voltage drop?
A few of the bypass cap choices seem a bit strange, mostly the epcos and os-con, not sure how I feel about tantalum.
Any reason for the 10 4.7µF film decoupling caps? Pricing in the US seems inflated as it is, why not get fewer but bigger caps instead? Concerned about high frequency AC rating on the 6.8µF ones? Pricing/Availability? Why not MKS4? Or simply because you started with them and then added more and more?
And a not capacitor related question: Are you aware of the NJM4556's DC offset drift with temperature and just don't consider it an issue with the power dissipation of the SIP8 package?
Don't get me wrong, I'm just trying to understand the design choices, not trying to talk it down.
Hi Setsul - all questions are welcome!
In both the ODA and NwAvGuy's O2 headamp the coupling caps in the middle (between the gain stage and the output current buffer) also form a high pass filter with the ground return resistor that has to be on the input of the current buffer stage to allow input bias current to flow.
In the case of the O2 the ground return resistor is 40.2K and his coupling cap is 2.2uF, so the corner frequency of the filter formed is 1/(2 * pi * 40.2K * 2.2uF) = 1.8Hz. In my version of the ODA I use a 1K pot to reduce Johnson noise (vs. the 10K in NwAvGuy's O2), and now it is looking like also reducing input impedance distortion (dScope measurements still pending by mgalusha). The ground return resistor has to be at least 5x the pot value (these resistors are across the pot wiper in both amps) to keep from loading the pot, so with my 1K pot that resistor reduces to 4.99K.
With 1/8 the ground return resistor value then in the ODA vs. the O2 the coupling cap value has to go up by at least 8x to maintain the same lower end corner frequency. My design goal with the ODA was to slightly exceed as many O2 parameters as I could, so ideally I wanted to drive that corner frequency even lower. Each ODA channel then has five 4.7uF caps in parallel for a total of 23.5uF. Another issue that figures in is the additive total capacitance tolerance with that many caps in parallel. In the worst case where all 5 caps were at their minimum end of the tolerance range I still wanted to exceed the O2's corner frequency (lower). The net result is an ODA corner frequency of 1 / (2 * pi * 4.99K * 23.5uF) = 1.39Hz. I won't do the math but as I recall if all 5 caps are at their minimum tolerance it is 1.6Hz or so, still lower than the O2's 1.8Hz.
Highly unlikely anyone could actually hear the difference between a low end frequency spectrum corner of 1.38Hz vs. 1.8Hz, but hey my design goal was to exceed the O2 and I met it! 😀
I was thinking about those 6.8uF WIMAs since I saw them in the data sheet, but last I looked Mouser (or anyone else I could find in the US) didn't stock them. They are also physically larger. I think I fiddled around with the numbers once and 4 of the 6.8's took about the same board space as 5 of the 4.7's. For what it is worth I notice there is an eBay vendor in Taiwan selling both the 4.7uF and those 6.8uF WIMAs (the 6.8s won't fit the board holes though) for less than Mouser. I always worry about fakes, but I see this fellow has photos of a wall full of OEM cases of the caps, so who knows, maybe they are legit.
And of course I couldn't use electrolytics for the coupling due to non-linearities and being a bipolar situation, so the only option was paralleling a bunch of film caps like this.
Those film caps are a significant part of the ODA's cost, about $50 USD all total. One way to reduce the cost with just a minimal ding to performance is use the 5K pot option in the BOM (either linear or audio taper, your choice, but the 1K has to be linear due to power dissipation). Then the ground return resistor can go up to 24.9K and the coupling caps only have to total up to one 4.7uF coupling cap on each channel. Just leave the other 4 coupling cap positions on each channel blank. NOTE: one of the 4.7uF WIMA caps on each channel isn't for coupling but is part of the DC offset zero circuit - check the schematic - those have to stay regardless. The downside is a slight increase in Johnson noise and probably input impedance distortion with the higher resistor pot, but saves a lot of money. There is also a small chance of an upside, slightly less distortion from the gain stage due to the lighter loading of the 5K pot vs. the 1K. That shouldn't be the case from the datasheet numbers - the LME49990 is THD plotted for a 600R load (pot plus feedback network) - but only the dScope tests can verify for certain. I've recently sent a message to Mike asking if this test could be done too. I have test points on the PC board on the output of each gain stage channel (different ground star segment though, the dScope ground should go to the input select switch body).
As for the decoupling caps, the tantalums are the datasheet preferred cap for those older LM317/337 parts because they require higher ESR to prevent oscillations. If I used a modern low ESR cap with those I would actually have to add a series resistor. But the LT3015/LT1963AA LDO final output regulators are just the opposite. They recommend low ESR, but not too low or it will oscillate. The organic polymer are a good fit here since they have low ESRs, but still have some measurable ESR unlike the ceramics (the datasheet goes into great detail here) which can have extremely low ESRs.
So to summarize, it is fitting in the best type of capacitor for the specific regulator, per the datasheets. 🙂
As for DC offset drift with the NJM4556ALs, the DC offset zeroing circuit takes their offset down from the 3.7mV in NwAvGuy's O2 amp to around 0.050mV = 50uV, or around 93% or so reduction. The amount of thermal drift I've actually measured over time has at most so far been around +/-100uV, which would still be over 90% better than the O2's DC output offset. That is one of the things that led to me settle on the fixed DC offset zero circuit vs. a servo.
In both the ODA and NwAvGuy's O2 headamp the coupling caps in the middle (between the gain stage and the output current buffer) also form a high pass filter with the ground return resistor that has to be on the input of the current buffer stage to allow input bias current to flow.
In the case of the O2 the ground return resistor is 40.2K and his coupling cap is 2.2uF, so the corner frequency of the filter formed is 1/(2 * pi * 40.2K * 2.2uF) = 1.8Hz. In my version of the ODA I use a 1K pot to reduce Johnson noise (vs. the 10K in NwAvGuy's O2), and now it is looking like also reducing input impedance distortion (dScope measurements still pending by mgalusha). The ground return resistor has to be at least 5x the pot value (these resistors are across the pot wiper in both amps) to keep from loading the pot, so with my 1K pot that resistor reduces to 4.99K.
With 1/8 the ground return resistor value then in the ODA vs. the O2 the coupling cap value has to go up by at least 8x to maintain the same lower end corner frequency. My design goal with the ODA was to slightly exceed as many O2 parameters as I could, so ideally I wanted to drive that corner frequency even lower. Each ODA channel then has five 4.7uF caps in parallel for a total of 23.5uF. Another issue that figures in is the additive total capacitance tolerance with that many caps in parallel. In the worst case where all 5 caps were at their minimum end of the tolerance range I still wanted to exceed the O2's corner frequency (lower). The net result is an ODA corner frequency of 1 / (2 * pi * 4.99K * 23.5uF) = 1.39Hz. I won't do the math but as I recall if all 5 caps are at their minimum tolerance it is 1.6Hz or so, still lower than the O2's 1.8Hz.
Highly unlikely anyone could actually hear the difference between a low end frequency spectrum corner of 1.38Hz vs. 1.8Hz, but hey my design goal was to exceed the O2 and I met it! 😀
I was thinking about those 6.8uF WIMAs since I saw them in the data sheet, but last I looked Mouser (or anyone else I could find in the US) didn't stock them. They are also physically larger. I think I fiddled around with the numbers once and 4 of the 6.8's took about the same board space as 5 of the 4.7's. For what it is worth I notice there is an eBay vendor in Taiwan selling both the 4.7uF and those 6.8uF WIMAs (the 6.8s won't fit the board holes though) for less than Mouser. I always worry about fakes, but I see this fellow has photos of a wall full of OEM cases of the caps, so who knows, maybe they are legit.
And of course I couldn't use electrolytics for the coupling due to non-linearities and being a bipolar situation, so the only option was paralleling a bunch of film caps like this.
Those film caps are a significant part of the ODA's cost, about $50 USD all total. One way to reduce the cost with just a minimal ding to performance is use the 5K pot option in the BOM (either linear or audio taper, your choice, but the 1K has to be linear due to power dissipation). Then the ground return resistor can go up to 24.9K and the coupling caps only have to total up to one 4.7uF coupling cap on each channel. Just leave the other 4 coupling cap positions on each channel blank. NOTE: one of the 4.7uF WIMA caps on each channel isn't for coupling but is part of the DC offset zero circuit - check the schematic - those have to stay regardless. The downside is a slight increase in Johnson noise and probably input impedance distortion with the higher resistor pot, but saves a lot of money. There is also a small chance of an upside, slightly less distortion from the gain stage due to the lighter loading of the 5K pot vs. the 1K. That shouldn't be the case from the datasheet numbers - the LME49990 is THD plotted for a 600R load (pot plus feedback network) - but only the dScope tests can verify for certain. I've recently sent a message to Mike asking if this test could be done too. I have test points on the PC board on the output of each gain stage channel (different ground star segment though, the dScope ground should go to the input select switch body).
As for the decoupling caps, the tantalums are the datasheet preferred cap for those older LM317/337 parts because they require higher ESR to prevent oscillations. If I used a modern low ESR cap with those I would actually have to add a series resistor. But the LT3015/LT1963AA LDO final output regulators are just the opposite. They recommend low ESR, but not too low or it will oscillate. The organic polymer are a good fit here since they have low ESRs, but still have some measurable ESR unlike the ceramics (the datasheet goes into great detail here) which can have extremely low ESRs.
So to summarize, it is fitting in the best type of capacitor for the specific regulator, per the datasheets. 🙂
As for DC offset drift with the NJM4556ALs, the DC offset zeroing circuit takes their offset down from the 3.7mV in NwAvGuy's O2 amp to around 0.050mV = 50uV, or around 93% or so reduction. The amount of thermal drift I've actually measured over time has at most so far been around +/-100uV, which would still be over 90% better than the O2's DC output offset. That is one of the things that led to me settle on the fixed DC offset zero circuit vs. a servo.
Last edited:
You misunderstood, I wasn't questioning the high-pass filter, the film caps for it or the total capacitance needed. I just think one big cap instead of 5 small ones would've been cheaper. Even if the MKS4 weren't available at that time too, what about this one:
http://www.mouser.de/ProductDetail/...=sGAEpiMZZMv1cc3ydrPrF4Mn0jw3NVER9Vpo1BlQ5hQ=
About the same size as 5 4.7µF WIMA MKS2 caps with spacing. Not that easy to find but I think 11$ instead of 36$ is enough of an incentive to at least try.
What about the Epcos and OS-CON for bypassing? The way you placed them inductance might cause more of a voltage drop than ESR. Cheaper and smaller caps closer to the pins maybe could've done the same.
Why Tantalum though? You choose Tantalum for lower ESR than Aluminium Electrolytics (although MLCC works too nowadays). If you want higher ESR than modern low ESR Elcaps why not get cheaper ones?
Did you actually calculate the maximum/minimum ESR before choosing the OS-CON?
http://www.mouser.de/ProductDetail/...=sGAEpiMZZMv1cc3ydrPrF4Mn0jw3NVER9Vpo1BlQ5hQ=
About the same size as 5 4.7µF WIMA MKS2 caps with spacing. Not that easy to find but I think 11$ instead of 36$ is enough of an incentive to at least try.
What about the Epcos and OS-CON for bypassing? The way you placed them inductance might cause more of a voltage drop than ESR. Cheaper and smaller caps closer to the pins maybe could've done the same.
Why Tantalum though? You choose Tantalum for lower ESR than Aluminium Electrolytics (although MLCC works too nowadays). If you want higher ESR than modern low ESR Elcaps why not get cheaper ones?
Did you actually calculate the maximum/minimum ESR before choosing the OS-CON?
I just checked, bulk price is 0.50$ in the US. Does only mouser sell them in smaller quantities or how can they charge 8 times that price?
If you can't get them anywhere else I could send you some so you can include them with the boards and drop the costs to 10$. I could also get the 0.22µF EPCOS for 0.05€, that would only save 3$ though.
If you can't get them anywhere else I could send you some so you can include them with the boards and drop the costs to 10$. I could also get the 0.22µF EPCOS for 0.05€, that would only save 3$ though.
Setsul - ah, I see. Sorry about the long winded answer! 🙂
The problem with those larger capacitance film caps is size. I did take a look at those before and they are 6mm longer than the current group of 4.7uF's. I couldn't get the layout to work out. With the pile of square 4.7uF's I can get them stuffed in the wider and less long area. You are quite right though - if the board was a bit bigger those longer caps would save a pile of money.
Yeah I've suspected that Mouser is making a lot of profit on those caps. They seem to be the only US distributor. Thanks for the offer at getting them cheaper. The at-cost group PCB buy here is done now but I probably would have taken you up on that at the time! 🙂
I'm having better luck with the 0.22uF Epcos. I found an overstock sale at Newark Electronics a couple of weeks ago for $0.078 each and bought 400! The page is here:
http://www.newark.com/epcos/b32529c224k/capacitor-pen-film-0-22uf-63v/dp/21C0562
On the bypassing, I typically keep the high(er) frequency bypass cap closer to the chip, while the parallel larger electro for lower frequencies I just try to keep in the vicinity. So for the output chips I have the two 560uFs centrally located with the 0.22uF's in between the chips. You are quite right about a larger number of smaller COG ceramics being a good idea, except in this case the bandwidth of those output chips is so low at 8mHz I didn't worry about it. If those had been LME49600s with their 110/180mHz GBP I most definitely would have ceramics right at each power lead. I've tried all sorts of reactive loads with the amp and so far haven't seen any oscillation at all on the scope.
On the tantalums, those are just on the voltage pre-regulator chips. I took the easy path and just stayed with the datasheet recommend for the tantalums. I tend to stick closely to the chip maker datasheet recommends since I know they have tested those scenarios. Since those are the pre-regulators they are pretty much out of the audio path. Their job is to reduce the ripple going into the final LDO regulators, plus in the case of the LT1963A with its 20V maximum input to drop the input voltage down. I did double the voltage rating on the tantalums though given their propensity to fail by going dead short and their voltage ratings being significantly de-rated with temperature.
The problem with those larger capacitance film caps is size. I did take a look at those before and they are 6mm longer than the current group of 4.7uF's. I couldn't get the layout to work out. With the pile of square 4.7uF's I can get them stuffed in the wider and less long area. You are quite right though - if the board was a bit bigger those longer caps would save a pile of money.
Yeah I've suspected that Mouser is making a lot of profit on those caps. They seem to be the only US distributor. Thanks for the offer at getting them cheaper. The at-cost group PCB buy here is done now but I probably would have taken you up on that at the time! 🙂
I'm having better luck with the 0.22uF Epcos. I found an overstock sale at Newark Electronics a couple of weeks ago for $0.078 each and bought 400! The page is here:
http://www.newark.com/epcos/b32529c224k/capacitor-pen-film-0-22uf-63v/dp/21C0562
On the bypassing, I typically keep the high(er) frequency bypass cap closer to the chip, while the parallel larger electro for lower frequencies I just try to keep in the vicinity. So for the output chips I have the two 560uFs centrally located with the 0.22uF's in between the chips. You are quite right about a larger number of smaller COG ceramics being a good idea, except in this case the bandwidth of those output chips is so low at 8mHz I didn't worry about it. If those had been LME49600s with their 110/180mHz GBP I most definitely would have ceramics right at each power lead. I've tried all sorts of reactive loads with the amp and so far haven't seen any oscillation at all on the scope.
On the tantalums, those are just on the voltage pre-regulator chips. I took the easy path and just stayed with the datasheet recommend for the tantalums. I tend to stick closely to the chip maker datasheet recommends since I know they have tested those scenarios. Since those are the pre-regulators they are pretty much out of the audio path. Their job is to reduce the ripple going into the final LDO regulators, plus in the case of the LT1963A with its 20V maximum input to drop the input voltage down. I did double the voltage rating on the tantalums though given their propensity to fail by going dead short and their voltage ratings being significantly de-rated with temperature.
Last edited:
Hello,
Just checking in I have been out of the loop for several weeks, been doing lots of home improvement projects!! LOL. Not as much fun as building electronics projects and listening to music.
My ODA has been used almost everyday now for several months! Always works well, I use the pre-amp output connected with RCA cables to an external AVR for computer sound and gaming or flying my X-Plane 10 simulator! (lol again).
I hardly ever use the O2 amps I have now.
I am waiting on the OPC USB dad etc.... hope its out soon.
All the best
Alex
Just checking in I have been out of the loop for several weeks, been doing lots of home improvement projects!! LOL. Not as much fun as building electronics projects and listening to music.
My ODA has been used almost everyday now for several months! Always works well, I use the pre-amp output connected with RCA cables to an external AVR for computer sound and gaming or flying my X-Plane 10 simulator! (lol again).
I hardly ever use the O2 amps I have now.
I am waiting on the OPC USB dad etc.... hope its out soon.
All the best
Alex
I finally got around to placing an order with Mouser. Most of the parts were in stock, but some very important components are backordered:
NRJ3HF-1 - 1/4" Stereo Jack - Estimated Ship: Dec. 17th, 2014
NJM4556AM - Op Amp - Mouser says: Will Advise
NJM4556AL - Op Amp - Estimated Ship: Feb. 19th, 2015
I can wait 2 weeks for the 1/4" jack, but what is the best way to deal with the NJM problem? Buy the DIP-8 package and adapt them? Anyone know if these are in stock somewhere else?
NRJ3HF-1 - 1/4" Stereo Jack - Estimated Ship: Dec. 17th, 2014
NJM4556AM - Op Amp - Mouser says: Will Advise
NJM4556AL - Op Amp - Estimated Ship: Feb. 19th, 2015
I can wait 2 weeks for the 1/4" jack, but what is the best way to deal with the NJM problem? Buy the DIP-8 package and adapt them? Anyone know if these are in stock somewhere else?
I have all those chips and the jack that I can sell you at-cost, plus whatever actual shipping is. 🙂 I'll send you a PM.
I've been meaning to post that the ODA power switch will no longer stocked at Mouser (they have 17 left), at least not without buying a thousand of them. I bought a bunch of them though when I saw this was going to happen, so if anyone needs a power switch after they run out just send me a PM.
I just ran a current stock check on the ODA BOM and here are a couple more...
The Alps version of the gain rotary switch is out of stock at Mouser, but I have a bunch of those too. Also in the notes column on the BOM the alternate Alpha switch is in stock, Mouser #105-SR10010F-24N. The only difference is the Alps shaft is 15mm long vs. 20mm long for the Alpha.
Mouser changed the way the 1N5249B-TR part number looks to 1N5249BTR and they are in stock.
The 5.49K leaded resistors are no longer stocked but I have a bunch of them. The 5.49K SMD resistors in the BOM are still in stock.
Any of these parts are at-cost, of course, for anyone on DIY audio building the ODA.
Last edited:
Typical...
Hi.
For C60 and C61 - could I use 47nF or 68nF as the 56nF tray is empty?
Regards
Hi.
For C60 and C61 - could I use 47nF or 68nF as the 56nF tray is empty?
Regards
An externally hosted image should be here but it was not working when we last tested it.
Hi Turbon,
C60 & C61 are there for an optional Zobel network and are not used. Same for R86 and R87. I have those in the "optional" price column on the BOM, but I really should add a note out in the BOM "notes" column too about it. I'll do that right now. 🙂
I left space for an output Zobel network just to cover all bases, in case oscillation showed with some specific loads, as happens with AMB's b22 occasionally. But never had a problem. I've tested it with all sorts of reactive loads and stable as a rock, same at the O2 with those NJM4556A chips.
C60 & C61 are there for an optional Zobel network and are not used. Same for R86 and R87. I have those in the "optional" price column on the BOM, but I really should add a note out in the BOM "notes" column too about it. I'll do that right now. 🙂
I left space for an output Zobel network just to cover all bases, in case oscillation showed with some specific loads, as happens with AMB's b22 occasionally. But never had a problem. I've tested it with all sorts of reactive loads and stable as a rock, same at the O2 with those NJM4556A chips.
Lol! I agree. 😀 I'm a bit OCD that way. Any gaps in the parts box must get taken care of! 😛 Newark Electronics is having a big overstock sale here. I just bulked up on all sorts of MLCC low-pf COG values for that very reason.
Last edited:
agdr, I see that you hae both 16VAC and 20VAC adapters in the BOM, is anyone to prefer over the other?
Regards
Regards
The 16Vac 2.4A adaptor is the preferred one for the US. 16Vac is all that is needed for the +/-12.5Vdc power rail build. 20Vac works but the extra just gets dissipated as heat. In your case you may need to put together a torroid in a separate box if you can't find an EU adaptor. I know there is a EU directive now about no more transformer wall warts, everything has to be switching DC supplies. In the case of a torroid the more common 18Vac secondary is just fine.
I used to think that these "travel converter" transformers that go from 240Vac to 120Vac would solve the problem, allowing use of the US WAU16-2400, but I forgot about the frequency difference. 50Hz stuff can run on 60Hz, but 60Hz stuff on 50Hz winds up with a 20% or so overvoltage.
I used to think that these "travel converter" transformers that go from 240Vac to 120Vac would solve the problem, allowing use of the US WAU16-2400, but I forgot about the frequency difference. 50Hz stuff can run on 60Hz, but 60Hz stuff on 50Hz winds up with a 20% or so overvoltage.
Thanks, a 18VAC 50VA toroid it will be.
Picking it up tomorrow.
A reflection - your build instuctions for v2.0 has been perfect so far (to the first test).
There is C48 I believe between C7 and C8 adjacent to R70 that appears in section 6 C44-C48 in the preamp section. I could be wrong but C48 (if it is) is easier to get in before C7/8.
Regards
Regards
Picking it up tomorrow.
A reflection - your build instuctions for v2.0 has been perfect so far (to the first test).
There is C48 I believe between C7 and C8 adjacent to R70 that appears in section 6 C44-C48 in the preamp section. I could be wrong but C48 (if it is) is easier to get in before C7/8.
Regards
Regards
Turbon - that 50VA sounds perfect! Thanks for the heads up on that instruction part sequence. I'll take a look at that.
Agdr, in section 1 - testing you wrote that the relay voltage should be 24vdc+-2vdc. Mine hoovers at about 26.4vdc. I haven't looked at it more closely but how would I lower it a bit? Maybe it solves itself when the toroid gets more loaded.
Regards
Regards
Hi Turbon - you are just fine with the 26.4V. The relay coil is good from +100% to -30% of the rated 24Vdc. On that particular sepcification I just pulled the 2% out of the air with the goal of checking that the voltage wasn't way off. I probably should make that 3% or 4%. 🙂
The fluctuation comes from that TL783 voltage regulator that does nothing but power the relay circuit. It requires a fairly hefty minimum load to stay within regulation. The datasheet says 15mA, but it turns out that is with the full 125V max input. There is a table in the datasheet which de-rates it for lower input levels. For what we have in the ODA the minimum load is around 7mA, which is where I put it with the bias resistor network. I kept that load near the minimum to reduce power dissipation in that chip. Once the relay turns on that load doubles and the voltage should get a a bit closer to the 24Vdc.
But long story short, even a +/-5Vdc variance in that 24V wouldn't matter at all. The only effect would be adding or chopping off a few tenths of a second to the relay delay time, due to the RC timing circuit.
The main LM317/337 and LT3015/LT1963A voltage regulators going out to the audio circuits though should be pretty much spot on though.
I'll change that in the build instrucitons to 4% during my next iteration. Thanks for the find! 🙂
The fluctuation comes from that TL783 voltage regulator that does nothing but power the relay circuit. It requires a fairly hefty minimum load to stay within regulation. The datasheet says 15mA, but it turns out that is with the full 125V max input. There is a table in the datasheet which de-rates it for lower input levels. For what we have in the ODA the minimum load is around 7mA, which is where I put it with the bias resistor network. I kept that load near the minimum to reduce power dissipation in that chip. Once the relay turns on that load doubles and the voltage should get a a bit closer to the 24Vdc.
But long story short, even a +/-5Vdc variance in that 24V wouldn't matter at all. The only effect would be adding or chopping off a few tenths of a second to the relay delay time, due to the RC timing circuit.
The main LM317/337 and LT3015/LT1963A voltage regulators going out to the audio circuits though should be pretty much spot on though.
I'll change that in the build instrucitons to 4% during my next iteration. Thanks for the find! 🙂
Last edited: