luvdunhill Posted June 28, 2009 Author Report Posted June 28, 2009 can you actually hear a difference when you flip the switch? I can. I hesitate to describe the difference as it will be forever recorded in the annals of time and all further impressions will either mirror mine or be polar opposite I'll wait for someone else to take the plunge first.
luvdunhill Posted June 28, 2009 Author Report Posted June 28, 2009 (edited) add a couple of extra resistors to make sure the switch does not temporarily disconnect all the feedback. That would be really cool, could you suggest something simple? I don't mind sacrificing this board once I finish with it to this test I think it would be a worthwhile addition to the boards. I figured a shorting switch would work, but maybe not? Edited June 28, 2009 by luvdunhill
pabbi1 Posted June 28, 2009 Report Posted June 28, 2009 That also works for me. Sounds sooooooooo nice with hd800's... Wish someone would allow me to buy connectors so that i could go balanced. (hint hint) Should be within the next two weeks...
luvdunhill Posted June 29, 2009 Author Report Posted June 29, 2009 Just had a probe slip and ended in a projectile BJT. That's a first. Just finished replacing the transistors and it's back up and running.
DigiPete Posted June 29, 2009 Report Posted June 29, 2009 poof... the magic smoke has been released...
Smeggy Posted June 29, 2009 Report Posted June 29, 2009 Crap, I was in Frys looking for long insulated probes like that but no joy, only those daft stubby things.
luvdunhill Posted June 29, 2009 Author Report Posted June 29, 2009 Ditto on Fry's. Guess I need to find something online. So, lowered the output resistors to 1 ohm and everything was stable. With the Dynahi current level, the current was shared exactly the same. Just for reference, the variance is 2mV on one side and 5mV on the other. With the insane output current setting (400mA per FET) things tightened up from a 17mV spread on one side, to 4mV and 66mV (my worse-case scenario) to 30mV difference. I think this is just about the limit of how hard I can both push the FETs and my heat sinks. With the 8 FETs, my targets will be 75mA and 200mA, and perform all the same measurements at these points. I gathered the data for dissipation on all the buffer transistors, just need to do some math. They still seem warm to me. Ideally, it would be nice if we could make sure the "faces" of the BJTs point away from each other, as this just exacerbates the heat... I think it would be a worthwhile change, albeit not simple I'd guess.
j4cbo Posted June 30, 2009 Report Posted June 30, 2009 How many of those heatsinks designed for the JFETs are available? We could retrofit some of those
n_maher Posted June 30, 2009 Report Posted June 30, 2009 My weapon of choice: I have those as well but I prefer to use these whenever possible. Those long fluke probes are hellaciously sharp
Looser101 Posted June 30, 2009 Report Posted June 30, 2009 Those long fluke probes are hellaciously sharp That's what I like about them. a) you know you are getting good contact since they don't slip, you can hold both probes with one hand and tweak with the other.
digger945 Posted July 1, 2009 Report Posted July 1, 2009 I like that better than holding a probe next to a resistor lead and accidentally piercing the solder mask and making contact with the ground plane (as your trying to hold both probes with one hand and adjust with the other). Be nice to have something that diggs in and wont slip. I will be looking for a pair of those R.
Smeggy Posted July 1, 2009 Report Posted July 1, 2009 Nothing more fun than popping a diode or two when stabbing through to ground
n_maher Posted July 1, 2009 Report Posted July 1, 2009 That's what I like about them. a) you know you are getting good contact since they don't slip, you can hold both probes with one hand and tweak with the other. But it also means they can puncture solder mask in the blink of an eye.
luvdunhill Posted July 4, 2009 Author Report Posted July 4, 2009 Well, I finished up testing the 8 FET version and mailed off the board to digger945. Once you get it Scott, can you do me one quick favor? Measure the values of the two "500 ohm" pots.. I think I'm going to try on my final build using a 470R resistor for one and a 510R resistor for the other, but want to verify which is higher value, as it's not clear from my notes. I think I'd like to remove the trim pots, and let the servo can keep things in check, per the original schematic. The final recommendation for the bias pot, is a 20K pot with 3K3 in series, as well as halving the two 20K resistors. In the final version, I used 1 ohm ballast resistors and they worked great. Truth be told, I would have liked to try 0.47R, but I think 1 ohm is good. I know the original target was 125mA per device. I think this is a good number, but felt like I preferred even more current, as the old addage that MOSFETs love current seems to be true. With the 8 FET configuration, the lowest bias you can really get, without further modifications (i.e. trial and error) to the circuit was 165mA per FET, given the values I used. If you don't have the heat sink to handle this, I'd recommend sticking with four FETs. Most people should plan on the four FET version, in my opinion. I maximum bias tried in the 8 FET configuration was 315mA per FET, for a total draw on the negative rail (where my ammeter was) of 1.3A. Things were rock stable, and in fact I didn't have the servo opamp installed. This thing does indeed scale quite nicely. I cannot wait to try it with speakers Anyways, good stuff. Let's start working on the modifications to the boards and get these off to production as soon as possible. I'll work on drawing up a final BOM by Monday and we can start firming up those numbers and working the logistics. Sooner we get cracking on this stuff, the sooner everyone can start building
digger945 Posted July 4, 2009 Report Posted July 4, 2009 That sounds good Marc. I feel confident that, with proper matching as you say, this can be built without the input bias pots, and the opamp can more than handle any small offset after warmup. Good show, I can't wait to get started.
DigiPete Posted July 4, 2009 Report Posted July 4, 2009 Cool!! The 4 fet per board is what I will build. What bias do you recommend in this config? I think 160ma per fet maximum would be accomodated by my twin dynahi psus (don't you think?)
luvdunhill Posted July 4, 2009 Author Report Posted July 4, 2009 That sounds good Marc. I feel confident that, with proper matching as you say, this can be built without the input bias pots, and the opamp can more than handle any small offset after warmup. Good show, I can't wait to get started. Well, I would recommend helping out the servo somewhat, otherwise it's a bit slow as it has to overcome the fact that the MOSFETs aren't so complimentary, so it would depend on your definition of warm up. From what I've seen, it looks like 5 minutes is about the minimum for things to really settle down and the bias to drop down from the turn-on peak to steady state. During this first 5 minutes, 200mA extra current isn't all that uncommon (worse case), before the negative tempco of the MOSFETs starts to work its magic. This will also depend on how large your heat sinks are, I'd presume. During the next 5 minutes after that, it will hit the point where it becomes rock solid, only varying around 10mA during this period. Anyways, I have written down 478R / 518R for one test and 478R / 523R for another. Worse case, just install one pot... I think with 470 / 510 it will work just fine.
luvdunhill Posted July 4, 2009 Author Report Posted July 4, 2009 Cool!! The 4 fet per board is what I will build. What bias do you recommend in this config? I think 160ma per fet maximum would be accomodated by my twin dynahi psus (don't you think?) I did testing specifically for your case. Just crank the pot all the way to it's maximum resistance (assuming that the series resistor I'm suggesting is added to the PCB) and pray that things don't get too hot From my testing, this will put you at nearly exactly the same current as the Dynahi output stage uses, and you should be fine. You'll have a bit more current in the PSU (you said you could dissipate a bit more heat here, so again, you should be fine), but the output stage should flow roughly the same. We can go from there.
digger945 Posted July 4, 2009 Report Posted July 4, 2009 Well, I would recommend helping out the servo somewhat, otherwise it's a bit slow as it has to overcome the fact that the MOSFETs aren't so complimentary, so it would depend on your definition of warm up. From what I've seen, it looks like 5 minutes is about the minimum for things to really settle down and the bias to drop down from the turn-on peak to steady state. During this first 5 minutes, 200mA extra current isn't all that uncommon (worse case), before the negative tempco of the MOSFETs starts to work its magic. This will also depend on how large your heat sinks are, I'd presume. During the next 5 minutes after that, it will hit the point where it becomes rock solid, only varying around 10mA during this period. Anyways, I have written down 478R / 518R for one test and 478R / 523R for another. Worse case, just install one pot... I think with 470 / 510 it will work just fine. That roughly parallels the time you really need for the Dynahi. The frequency of the filter for the opamp integrator is 0.05Hz so yea it's pretty slow to react to input. As you mentioned it would be wise for anyone using this design to allow some time for things to settle in as they warm up. On a side note, Jacob has stated to me that the Uber controller may have input offset monitoring incorporated into the design.
luvdunhill Posted July 4, 2009 Author Report Posted July 4, 2009 (edited) Wrt Conrad MF30-75's(@3"x12") I have 16 outputs at 75mA each per HS and it's only just warm I think that the MF30-100 is a solid recommendation for the DynaFET at this point. This is based on the results of my heat sink and some crude math. This heat sink is 300mm wide, 100mm tall with 30 48mm fins. The base is 8mm thick. It is rated as 0.32 C/W @ 80C, or around 0.35 @ 60C. That results in 171W at @60C. Given two boards of 4 devices and 30V rails, that right at Kevin's recommendation of 150mA, at a conservative die temperature. I wouldn't even recommend using 8 devices with these. This is my standard recommendation, if anyone is looking for one. If you're using another heat sink that you don't trust the C/W rating for one reason or another, you can use a crude method of comparing the two based on dimensions alone. Edited July 4, 2009 by luvdunhill
digger945 Posted July 4, 2009 Report Posted July 4, 2009 The Conrad MF30-100 would work great with a Parmetal 12x12x4" enclosure.
digger945 Posted July 7, 2009 Report Posted July 7, 2009 Just now recieved the board. I will mount it to my heatsink tomorrow. How thick was your L bracket Marc?
kevin gilmore Posted July 8, 2009 Report Posted July 8, 2009 I'm about 95% done stuffing the board marc sent me, this is definitely NOT a beginner project. I'll be machining up some custom brackets in the next couple of days. If anyone sells boards to -=her=- i will hunt them down and kill them. Seriously a person with her skills has no chance at this thing.
Icarium Posted July 8, 2009 Report Posted July 8, 2009 Wha? Has news of this project leaked to her? If so.. death! Death!!!
Recommended Posts
Create an account or sign in to comment
You need to be a member in order to leave a comment
Create an account
Sign up for a new account in our community. It's easy!
Register a new accountSign in
Already have an account? Sign in here.
Sign In Now