Thanks again the type of output stage i am refering to is http://patft.uspto.gov/netacgi/nph-...0&s1=4833424.WKU.&OS=PN/4833424&RS=PN/4833424
http://patft.uspto.gov/netacgi/nph-...0&s1=6600367.WKU.&OS=PN/6600367&RS=PN/6600367
and our own Nelson Pass http://patft.uspto.gov/netacgi/nph-...0&s1=3995228.WKU.&OS=PN/3995228&RS=PN/3995228
http://patft.uspto.gov/netacgi/nph-...0&s1=6600367.WKU.&OS=PN/6600367&RS=PN/6600367
and our own Nelson Pass http://patft.uspto.gov/netacgi/nph-...0&s1=3995228.WKU.&OS=PN/3995228&RS=PN/3995228
You are not playing fair, John........
Do you really expect our resident "tyrannical librarian" to be familiar with an H-P Journal (from an evil American company, no less!) from 1971? Probably before his time...........
Anyway......I have used that relationship when building the type of output stage that you show in post #19, with multiple devices. Saw no reason not to.
Jocko
Do you really expect our resident "tyrannical librarian" to be familiar with an H-P Journal (from an evil American company, no less!) from 1971? Probably before his time...........
Anyway......I have used that relationship when building the type of output stage that you show in post #19, with multiple devices. Saw no reason not to.
Jocko
I did not come up with that expression.......
Traderbam did.
http://www.diyaudio.com/forums/showthread.php?postid=483948#post483948
No.......haven't seen or talked to Mr. Personality recently.
Jocko, the Other Mr. Personality
Traderbam did.
http://www.diyaudio.com/forums/showthread.php?postid=483948#post483948
No.......haven't seen or talked to Mr. Personality recently.
Jocko, the Other Mr. Personality
Yup, you're right. Better than be believing is measuring. Just build a test bench with a two stage EF and a single NPN/PNP output pair (MJL1302/MJL3281). I reduced the emitter resistors to 0 (zero) ohms and adjusted the bias to 100ma.
Astoundingly this circuit is thermal stable, much more thermal stable than expected. After heating up to about 60°C the bias changed slightly to 150ma. And I can't see any spikes in the error signal. Made a quick FFT with my scope, only K2 is present. No K3 and higher orders, but the dynamic range of my scope is limited to 60 dB only, so this FFT is not very meaningful.
Hi, bocka,
Without emitter resistors there wouldn't be any peculiarities at low input voltages. You should observe the difference between input and output signal, the output signal should be attenuated to eliminate linear part from the error signal. In that way you will increase the dynamic range of your measurement system by several orders.
Without emitter resistors there wouldn't be any peculiarities at low input voltages. You should observe the difference between input and output signal, the output signal should be attenuated to eliminate linear part from the error signal. In that way you will increase the dynamic range of your measurement system by several orders.
Bocka,
Right on. I don't have enough equipment for a lot of measuring I would like to do, so often have to rely on what otheres have posted/published. Nonetheless, I try to discern whether the opions are from personal measurement, measurement reported elsewhere, theoretical analysis or conventional wisdom (belief).
Ultimately, actually building something is the test. I have found some great sounding ideas that are not so great when constructed (at least without resorting to heroic efforts) and in my ignorance, I've constructed a couple of things that perform just fine despite assurances that they won't.
Anyway, good for you for actually building as opossed to just opining.
Hi dimitri,
I did this yesterday later. The error signal is 2 magnitudes below the output signal and does not show any spikes just low order distortion up the the six order (with 2 ohms load resistance). This is because every BJT has an intrinsic emitter resistance. Without emitter degeneration and very low voltage drop over small valued emitter resistors distortions drops in the way my simulation shows.
I want to come back to my initial topic:
Assume that I want to build a power amp with about 150W/8ohms. This will result in a power supply voltage of about +/- 60V. For that power two heatsinks of about 0.5K/W should be sufficiant one for the PNPs and the other one for the NPNs. With 15K temerature rising with no load this results in about 0.5A idling current. When paralleling 6 devices each device is biased at about 80ma. To me the question is:
Wich emitter resistor is best?
Well, it can calculated with Re = 0.5 .. 1 * Vkt / Ibias, with Vkt = 30mV (not 26mV which value is most given because Tj is not 300K but more 320 .. 330K). So I come to Re = 0.19 .. 0.38 ohms. For highest thermal stability 0.38 ohms shold be best. So, carefully designed, I'd take a 0.39 ohm remitter resistor and bias Iidle to that point, where lowest distortion occurs.
But, when simulate this, I obtain 0.12 ohms is best with lowest distortion, better with lower resistance and quite independent from idling current. It makes nearly no difference if the idling current is 60mA, 80mA or 120mA. This is not covered by theory with much too simplified models where Is(PNP) = Is(NPN) all other parameters also equal as well as the therory only is done with a single EF, not with dual or triple darlingtons.
I'm shure that there is a better way designing ouput stages. And this way I found by simulation is that lowering the emitter resistors can improve the quality of the output stange. And that the theory about transconductance doubling suffers froms too simplified models.
Right. But there are also differences from human beings because the interpretation from measurement differs. So a discussion is worthfull. And - maybe someone won't believe it - I strongly believe in theory, but only in that way, where it covers reality, not just with simplified theoretical models. Most of electronic design can be done with simple calulations and a lot of experience. Sometimes not, if you want to design an analog multiplier you need mathematics and the theory how a BJT works. And practical experience, why this can be done in a monolytic device an will not work well with descrete transistors and why you need thermal compensation.
The attached Schematic contains Design Enhancements that our esteemed members like Walt Jung Multi-loop Feedback technique to maintain a constant feedback factor across the audio range and a Constant 0.0033% THD and 0.02% IMD @ 1V into a Real 24 ohm Headphone load from the Sony MDR-V700 Dj Phones. And also gets about 0.003% THD & 0.005% IMD (SMPT) at 1 V into a similar 24 ohm resistor. Reactive Loads I guess. Anyway Thanks Walt this Technique works extremely well.
Nelson pass is Responsible for my Investigation into Dynamic Biasing so what better place to test all the different ways to lingerie the output stage as much as possible since the Multi-loop topology reduces the available feedback at low frequencies so distortion is higher however constant with frequency. I remember hearing something in sockeyes work abut the 20 MV things but search did not turn up anything. Then at the right time comes John with the perfect answer. Thanks john for contributing to my enjoyment of DIY projects. The Diode Trick was a real neat answer as if I adjuster the bias tow that low the input transistors would turn off at the onset of clipping si at 130 mV this I assumed was causing more than messier attention in the crossover reason. However since one of the purported claims of dynamic biasing is never allowing the non conducting transistor to become reversed biased? This nationally just made my pass on this is issue as both the sound and measured performance was varying well.
Well after countless instability issues at low Bias currents I redid the resistors to give 20 mV at 20 mA Icq. This resulted in the same THD and IMD as the normal Diamond buffer built with the same parts I was previously using. Since the output stage is a plug in module a quick comparison of several slightly versions could be evaluated on the same platform. One thing I noticed is that while THD and IMD were no better in the first and I must admit rather sloppy implementation of Dynamic Biasing is that THD and IMD remained constant with loads from 24-300 ohms. It’s like identical over the entire 24-300 Ohm range; also Distortion measured same Resistor or real headphone load. This I thought was impressive since the standard diamond buffer changed over a three to one range decreasing at higher impedances. Well then I tried the diode trick and THD and IMD was cult almost in half and the sound regained some its dynamics in comparison with just the 10 ohm Emitter resistors that was 4.7 ohms on the Previous Diamond buffer.
So thanks Mr. Pass for introducing yet another creative twist to Audio design
Nelson pass is Responsible for my Investigation into Dynamic Biasing so what better place to test all the different ways to lingerie the output stage as much as possible since the Multi-loop topology reduces the available feedback at low frequencies so distortion is higher however constant with frequency. I remember hearing something in sockeyes work abut the 20 MV things but search did not turn up anything. Then at the right time comes John with the perfect answer. Thanks john for contributing to my enjoyment of DIY projects. The Diode Trick was a real neat answer as if I adjuster the bias tow that low the input transistors would turn off at the onset of clipping si at 130 mV this I assumed was causing more than messier attention in the crossover reason. However since one of the purported claims of dynamic biasing is never allowing the non conducting transistor to become reversed biased? This nationally just made my pass on this is issue as both the sound and measured performance was varying well.
Well after countless instability issues at low Bias currents I redid the resistors to give 20 mV at 20 mA Icq. This resulted in the same THD and IMD as the normal Diamond buffer built with the same parts I was previously using. Since the output stage is a plug in module a quick comparison of several slightly versions could be evaluated on the same platform. One thing I noticed is that while THD and IMD were no better in the first and I must admit rather sloppy implementation of Dynamic Biasing is that THD and IMD remained constant with loads from 24-300 ohms. It’s like identical over the entire 24-300 Ohm range; also Distortion measured same Resistor or real headphone load. This I thought was impressive since the standard diamond buffer changed over a three to one range decreasing at higher impedances. Well then I tried the diode trick and THD and IMD was cult almost in half and the sound regained some its dynamics in comparison with just the 10 ohm Emitter resistors that was 4.7 ohms on the Previous Diamond buffer.
So thanks Mr. Pass for introducing yet another creative twist to Audio design
Hi bocka !
According to your measurements, does this mean that the
mjl3281/1302 are thermal very stable ? This could mean
that with a single pair an amp with these could be built without
Re's ? And that i might place the bjt for the Vbe-multiplier
far from the heatsink ?
If they are that stable, this would explain why my amp is thermally
overcompensated...
MikeB
According to your measurements, does this mean that the
mjl3281/1302 are thermal very stable ? This could mean
that with a single pair an amp with these could be built without
Re's ? And that i might place the bjt for the Vbe-multiplier
far from the heatsink ?
If they are that stable, this would explain why my amp is thermally
overcompensated...
MikeB
Hi MikeB!
No my measurements only shows that my EF is stable. I took a short wiew in the MJL1302/3281 datashee (and a simple calculation) and found that the MJL1302 has an intrinsic emitter resistance of about 0.045 ohms and the MJL3281 of 0.12 ohms. This may explain why I see no crossover distortion.
Yes, I think so.
No! The BJTs must thermally coupled in the best way to prevent thermal runaway. I placed my Vbe-multipier directly between the MJL1302/3281 pair. I think your amp is thermally overcompensated because the dVbe/dT of your different devices are not mached.
No my measurements only shows that my EF is stable. I took a short wiew in the MJL1302/3281 datashee (and a simple calculation) and found that the MJL1302 has an intrinsic emitter resistance of about 0.045 ohms and the MJL3281 of 0.12 ohms. This may explain why I see no crossover distortion.
Yes, I think so.
No! The BJTs must thermally coupled in the best way to prevent thermal runaway. I placed my Vbe-multipier directly between the MJL1302/3281 pair. I think your amp is thermally overcompensated because the dVbe/dT of your different devices are not mached.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.