I will have boards made up -- a few hundred -- there are, however, about 4 configurations to twiddle with -- I am intrigued with the ThermalTrak guys from On-Semi, Lateral MOSFET's, the Sanken devices -- as per usual I don't stray from NatSemi's recommendations -- thank you, no minimalist here -- if you want to leave parts out, do so at your discretion.
I've been working with my Beveled Thors http://www.tech-diy.com/Loudspeakers/NorseDesigns/Beveled_Thor_Painted.jpg which has taken a lot of time -- the paint has finally cured -- so it's on to LM4702.
jack
I've been working with my Beveled Thors http://www.tech-diy.com/Loudspeakers/NorseDesigns/Beveled_Thor_Painted.jpg which has taken a lot of time -- the paint has finally cured -- so it's on to LM4702.
jack
jackinnj said:
Thanks. I'd be interested in purchasing some from you. Will the boards work both LM4702C and LM4702B and possibly the LM4702A, if it goes into production? Will these boards be available this year? I'm in no hurry, here. Just asking for an approximation, that's all.
Thanks,
David
I have had some prototypes made up -- I found that the compensation caps are better off as silver mikes and not ceramics, some of the caps are a bit larger, National used some fusible resistors in the zobels -- the proto's with the Sanken devices got down to 0.01% THD which is far short of what can be achieved.
I don't think that the TO-3 style device is practical for DIY -- these devices are almost impossible to remove if soldered in -- will stick with the Pentawatt devices.
I don't think that the TO-3 style device is practical for DIY -- these devices are almost impossible to remove if soldered in -- will stick with the Pentawatt devices.
Hey, count me in on boards. I am working on some too but a little new to this discrete like design so would like to learn from others.
Seems paralleling Darlingtons is not good for stability. I want to work on a design with no darlingtons by breaking out the darlingtons into discretes then see if I can get several paralleled up for higher power, especially now that the B version it out. Anyone look into using Sziklais for this? I haven't yet but wonder if it will work as well.
Thx. Still waiting on Verteran to post back his results using the LM4702 and FETs plus get in on a group buy from him.
-SL
Seems paralleling Darlingtons is not good for stability. I want to work on a design with no darlingtons by breaking out the darlingtons into discretes then see if I can get several paralleled up for higher power, especially now that the B version it out. Anyone look into using Sziklais for this? I haven't yet but wonder if it will work as well.
Thx. Still waiting on Verteran to post back his results using the LM4702 and FETs plus get in on a group buy from him.
-SL
I've gotten the THD down to 0.002% at 1kHz -- here's the work in process:
An externally hosted image should be here but it was not working when we last tested it.
I've gotten a FET version working on the bench. It is basically Verteran's design which is just a very simple change from using Darlingtons or BJTs (add in a 10 ohm gate resistor). I was worried about fire and smoke so have been taking it very slow. THD+N measurements are not great, like 0.02% (80kHz BW) at 1kHz. THD+N slopes up some with frequency. I think maybe I have a small oscillation somewhere. Gotta try to optimize for better THD and noise. Listening to the sound I thought it sounded very clean with great definition of all instruments. Not a blind test and nothing to compare it too so pretty subjective. This is my first working discreteamp, if we can call it that.
BTW, for info it seems setting the Vce voltage on the Vbe multipier to 3.0 - 3.1V works pretty well and keeps bias current reasonsable. Running FFTs there was some harmonics reduction with higher vbe voltage but bias current went up significantly. Going lower didn't reduce bias current that much at the expense of higher harmonics in the FFT plot.
Fun stuff, now to get some time to try and improve performance. Also need to test with an unregulated supply instead of bench supplies since that is how it will be used.
-SL
BTW, for info it seems setting the Vce voltage on the Vbe multipier to 3.0 - 3.1V works pretty well and keeps bias current reasonsable. Running FFTs there was some harmonics reduction with higher vbe voltage but bias current went up significantly. Going lower didn't reduce bias current that much at the expense of higher harmonics in the FFT plot.
Fun stuff, now to get some time to try and improve performance. Also need to test with an unregulated supply instead of bench supplies since that is how it will be used.
-SL
SpittinLLama, You are first person outside Poland that assembled this amplifier on my PCB. How is this thing working? What transistors are you using on output? Some pictures? LM4702 is quite sensitive device.
I have one more version of PCB ready - including microcontroller protection circuits (DC, thermal, over-current, soft-start).
I have one more version of PCB ready - including microcontroller protection circuits (DC, thermal, over-current, soft-start).
SpittinLLama said:
What type output devices are you using -- I am using Lateral MOSFETs --
Try putting the gate stopper resistor right on the transistor -- and use the belt and suspenders low pass filtration.
I agree with Veteran -- this is a very sensitive device -- just moving the power supply leads about will lead to differeing THD measurements. Of course, the LM3875, LM3886 and LM4780 have the same wrinkles.
I'm using a BD139 for the Vbe multiplier. The pot is a 1K to give me some range to adjust the voltage with 1.2K ohms at top of the vmult section. The snubber is 10ohms + 0.1uF (or 100nF is you prefer). I am using 0.12 ohms for the source resistors and 10 ohms for the gate resistor. It is pretty much Verteran's schematic with missing values filled it. Because the vbe voltage is not real high the FETs are lower Vt ones. 2SK1530 & 2SJ201 for outputs, as mentioned by Verteran in one of his post. This really is a simple device to use, as shown in the datasheet. Just the devil is in the details when it comes to performance.
It is sensitive. Moving my input wires will change THD. My vbe voltage drifts with temperature but it should, right?
It works very well. I had it as high as +/-45V (only had 50V caps around so keeping it safe) and can run 100W into 8 ohms with this set up. I could try to push it harder since the FETs are 12A rated but my bench supplies are only rated 60V/9A. I don't have any ability to push it real hard. I would like to try paralleling up some more FETs but since I can't really test running hard anyway, not sure it is worth it. Adding more output devices should be pretty simple.
I have no protection. I am looking at other examples. Pretty much short circuit and DC voltage should cover me. Thermal protection and others would be nice but not as important to me.
JackinNJ, not sure what you mean. The gate resistor is about as close as possible to the FETs as a through hole can be. The only filtering is the snubber and then the 30pF compensation caps, like the datasheet shows. Are you referring to some additional things? Like I said, I think there is a small high oscillation. Oh, and if I don't connect the inputs of the LM4702 to something the output of the amp oscillates.
-SL
It is sensitive. Moving my input wires will change THD. My vbe voltage drifts with temperature but it should, right?
It works very well. I had it as high as +/-45V (only had 50V caps around so keeping it safe) and can run 100W into 8 ohms with this set up. I could try to push it harder since the FETs are 12A rated but my bench supplies are only rated 60V/9A. I don't have any ability to push it real hard. I would like to try paralleling up some more FETs but since I can't really test running hard anyway, not sure it is worth it. Adding more output devices should be pretty simple.
I have no protection. I am looking at other examples. Pretty much short circuit and DC voltage should cover me. Thermal protection and others would be nice but not as important to me.
JackinNJ, not sure what you mean. The gate resistor is about as close as possible to the FETs as a through hole can be. The only filtering is the snubber and then the 30pF compensation caps, like the datasheet shows. Are you referring to some additional things? Like I said, I think there is a small high oscillation. Oh, and if I don't connect the inputs of the LM4702 to something the output of the amp oscillates.
-SL
SL -- I had a 5.6 MegaHertz !!! Oscillation in the first LM4702 I built with the Sanken devices -- I also thought I had a 40kHz problem but it turned out to be a new fluourescent lamp which I began using a few weeks ago. Talk about "stuck on stupid" 
Memo to file -- don't buy your lamps at Home Depot.
Memo to file -- don't buy your lamps at Home Depot.
SpittinLLama said:
You can use 3.3k resistor ‘at top’. This should give you full range of voltage (a little more then 6V). 100mA of stand-by current for each channel.
When temperature is rising Vbe voltage should fall down.
I have microcontroller based protection circuit: DC, thermal, transformer soft start, over-current. This will be a part of my new board.
jackinnj said:SL -- I had a 5.6 MegaHertz !!! Oscillation in the first LM4702 I built with the Sanken devices -- I also thought I had a 40kHz problem but it turned out to be a new fluourescent lamp which I began using a few weeks ago. Talk about "stuck on stupid"
Memo to file -- don't buy your lamps at Home Depot.
I placed few more small caps (56pF-1000pF) – input shunt resistor, feedback and I think that there should be a preamplifier for this amp – input is quite unstable and sensitive.
One more thing – SpittinLLama, if you would like to use more output transistors remember, that LM4702 output current is quite small and connecting transistors in parallel will increase gate capacitance.
that optocoupler has a temperature coefficient -- and they aren't particularly linear anwyay (unless you use one with a servo like the HCNR201 or IL300) -- why not try this -- use a TL431 as a comparator -- it won't load the emitter/source resistor -- yeah, I know the TL431 is a shunt regulator but it can also be used as a comparator with its own reference -- and it is somewhat tempco stabilized. See the Texas Instruments product folder.
you can use then TL431 (which costs $0.30) to drive the optocoupler.
you can use then TL431 (which costs $0.30) to drive the optocoupler.
Thanks for all the inputs, guys.
Veteran, You added caps in several places I plan on adding them. At the input I typically use 15pF - 47pF. The one from base to collector of the Vbe multiplyer was one too but I hadn't decided on a value yet, just small. I was also going to try a small film cap across the collector - emitter in parallel with the 10uF one to see if any improvement could be discerned. Another cap on the trim pot might be helpful so gonna try that too, something like a 0.1uF. I am going to check out adding one from Gate to Source on one of the FETs so the gate capacitance is close to equal between the two FETs. I saw this on a schematic somewhere and thought I'd try it out. To be honest, I don't know why this would make any difference but would like to try it and see for myself.
I don't understand why there needs to be source resistors (0.1 or 0.22 like you have now) when running just one set of FETs if there is no protection circuitry. Seems like the Rds(on) of the FETs will serve mostly the same purpose. Am I confused again? If paralleling FETs then I think they would be there to be conservative but maybe not even necessary then? I know and understand why they are needed for BJTs but not FETs. Since a newbie I might be confused.
Thanks for the clarification on the vbe voltage. At first I was like, what is going on then I thouight about it and realized that is what should be happening. But for driving FETs is this good, bad, or no issue? For driving BJTs it is a good thing but doesn't seem like you would want it to move for driving FETs. Another newbie type question.
Why the change in gate resistor value? I was going to ask if there are any sources on getting this value optimized or if the only way was through trial and error on the bench. Maybe this value is not critical at all. Going from 10 to 220 ohms is a significant difference in value.
I think I am fine with the Vbe Multiplyer the way it is now. I have enough range for setting bias current. It runs about 50mA from each bench supply. Going higher didn't seem to have a big effect on harmonics but maybe I will change and re-test.
I had a slight bump in the THD vs output power plot as it neared clipping. I tried adding a 1uF film cap across Vcc - Vee right at the board terminals and the bump is now gone. The plot now goes smoothly down until clipping and then shoots up, just like it should. Might be something you want to add space for on the board.
I think a pre-amp is a good idea so gain on the LM4702 can be lowered some. I might lower the gain just to see how the performance changes. At the very least, good information to have.
I hope to get some time today to try some of these ideas out. I was also going to post the bump plot but seems time is hard to come by.
Thanks for all the info.
-SL
Veteran, You added caps in several places I plan on adding them. At the input I typically use 15pF - 47pF. The one from base to collector of the Vbe multiplyer was one too but I hadn't decided on a value yet, just small. I was also going to try a small film cap across the collector - emitter in parallel with the 10uF one to see if any improvement could be discerned. Another cap on the trim pot might be helpful so gonna try that too, something like a 0.1uF. I am going to check out adding one from Gate to Source on one of the FETs so the gate capacitance is close to equal between the two FETs. I saw this on a schematic somewhere and thought I'd try it out. To be honest, I don't know why this would make any difference but would like to try it and see for myself.
I don't understand why there needs to be source resistors (0.1 or 0.22 like you have now) when running just one set of FETs if there is no protection circuitry. Seems like the Rds(on) of the FETs will serve mostly the same purpose. Am I confused again? If paralleling FETs then I think they would be there to be conservative but maybe not even necessary then? I know and understand why they are needed for BJTs but not FETs. Since a newbie I might be confused.
Thanks for the clarification on the vbe voltage. At first I was like, what is going on then I thouight about it and realized that is what should be happening. But for driving FETs is this good, bad, or no issue? For driving BJTs it is a good thing but doesn't seem like you would want it to move for driving FETs. Another newbie type question.
Why the change in gate resistor value? I was going to ask if there are any sources on getting this value optimized or if the only way was through trial and error on the bench. Maybe this value is not critical at all. Going from 10 to 220 ohms is a significant difference in value.
I think I am fine with the Vbe Multiplyer the way it is now. I have enough range for setting bias current. It runs about 50mA from each bench supply. Going higher didn't seem to have a big effect on harmonics but maybe I will change and re-test.
I had a slight bump in the THD vs output power plot as it neared clipping. I tried adding a 1uF film cap across Vcc - Vee right at the board terminals and the bump is now gone. The plot now goes smoothly down until clipping and then shoots up, just like it should. Might be something you want to add space for on the board.
I think a pre-amp is a good idea so gain on the LM4702 can be lowered some. I might lower the gain just to see how the performance changes. At the very least, good information to have.
I hope to get some time today to try some of these ideas out. I was also going to post the bump plot but seems time is hard to come by.
Thanks for all the info.
-SL
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.