I notice there are a lot of darlington power transistor that are TO-3P and TO-264 with the same pin out as the most common transistors. Is there any reason people don't use darlington? You guys keep talking about 3EF, why not use darlington and a pre-driver to get 3EF? Or you can upgrade existing amp that is 2EF with a darlington? All you need to do is to adjust the Vbe multiplier to 6Vbe.
Integrated Darlingtons aren't that great for high-end power amps. The reasons are:
You can't get at the emitter resistor of the driver, so your options for driver bias are limited.
They are slower and have worse SOA than a "discrete Darlington" (double EF) made of two modern high performance BJTs like the MJE15032 and MJL3281. This is partly because of the driver bias issue and partly because the Darlingtons are made on an older, slower process. (The popular TIP142/147 pair have a Ft of about 4MHz, the MJE/MJL discrete option would give more like 40MHz.)
I personally prefer the EF over the Sziklai pair because it has a broader and smoother crossover region. Having said that, I guess the Sziklai pair can be smoothed out, that is what the Baxandall diode does.
You can't get at the emitter resistor of the driver, so your options for driver bias are limited.
They are slower and have worse SOA than a "discrete Darlington" (double EF) made of two modern high performance BJTs like the MJE15032 and MJL3281. This is partly because of the driver bias issue and partly because the Darlingtons are made on an older, slower process. (The popular TIP142/147 pair have a Ft of about 4MHz, the MJE/MJL discrete option would give more like 40MHz.)
I personally prefer the EF over the Sziklai pair because it has a broader and smoother crossover region. Having said that, I guess the Sziklai pair can be smoothed out, that is what the Baxandall diode does.
Last edited:
Member
Joined 2009
Paid Member
I have a pot full of Sanken power Darlington's taken from a Pioneer HT receiver. They look like pretty nice devices to me. I'd be quite willing to use them in an amplifier design.
I believe Gabobela on this forum has had some good experience with power Darlington devices, his favourite DIY amp of 20yrs is based on them.
I would probably shy away from using them in a design with parallel outputs for high power as I don't expect it would be as easy to match them in pairs - although it would be interesting to measure those that I have.
A 3EF has more phase shift etc than a 2EF so 'upgrading' from one to the other might require some additional components for overall stability of the amplifier.
I believe Gabobela on this forum has had some good experience with power Darlington devices, his favourite DIY amp of 20yrs is based on them.
I would probably shy away from using them in a design with parallel outputs for high power as I don't expect it would be as easy to match them in pairs - although it would be interesting to measure those that I have.
A 3EF has more phase shift etc than a 2EF so 'upgrading' from one to the other might require some additional components for overall stability of the amplifier.
It's my biased belief that Darlingtons were invented by "production engineers" to reduce the assembly costs of amplifiers.
I can't see any use, where a two transistor assembly would perform worse than an integrated Darlington. In most cases the two transistor will perform better than the integrated Darlington.
I can't see any use, where a two transistor assembly would perform worse than an integrated Darlington. In most cases the two transistor will perform better than the integrated Darlington.
Darlingtons don't mix PNP and NPN, so quasi amps are out, and usually PNP driver gives lower PS overhead.
usually designed for selling lousy output devices paired with any old driver to give new super gain specs.
usually designed for selling lousy output devices paired with any old driver to give new super gain specs.
Last edited:
Please take another look; the "Baxandall diode" seems to be a way of improving the crossover distortion of a quasi-complementary output stage (NPN pullup, NPN pulldown with Baxandall diode). This circuit was invented in The Old Days, when you couldn't buy a high-beta, high-power, high-fT PNP transistor for love or money.
An output stage made of 1 Sziklai pullup (NPN+PNP) and 1 Sziklai pulldown (PNP+NPN) has no need of a Baxandall diode.
See, for example, Douglas Self's "Trimodal Amplifier" (Fig 17.19 of APADH 6th ed). TR6-TR7 are a Sziklai pair which pulls the output up towards +Vcc, and TR8-TR9 are another Sziklai pair which pulls the output down towards -Vee. There is no Baxandall diode.
Edit- oh, almost forgot: Here is a 100 watt audio amplifier designed by Nelson Pass which uses Darlington transistors in its output stage. (The very inexpensive TIP142 and TIP147). Mr. Pass's schematic included rather nonstandard notation for the resistor values; a subsequent poster has kindly provided a translation table.
Last edited:
Lets not forget Nelson's A40 that used Lambda darlingtons because of their low value driver emitter resistor.
Craig
Craig
Member
Joined 2009
Paid Member
I used Sanken power Darlington outputs in my TGM6 amplifier. It had an LTP input stage with a Sziklai VAS driving the outputs. It was configured for fairly high feedback factor. I simulated it using Spice on the basis of a discrete Darlington but with the same emitter resistor as in the integrated devices and was able to obtain a good simulation result. On building the amplifier I found that the sound was good, the final version was not far from my old Bryston amplifier in terms of a strong clean sound. It was not as good as my TGM8 (and I scrapped it due to issues unrelated to the amplifier), but it was certainly hi-fi. Hence I have kept my stock of these Sanken devices as I may well use them again in another project.
I'm not sure how much cost benefit was obtained by the use of power Darlingtons in my Pioneer amplifier, they could have used cheaper parts. It doesn't even use SMT transistors at the front end which would be cheaper too. I've opened up really cheap amplifiers and they didn't use power Darlingtons. Pioneer have their reasons but I"m not sure what they are.
Certainly, there is more flexibility when you build a discrete Darlington and there are some really superb hi-end devices available, especially the older Japanese parts. I always assumed that integrated Darlington devices were 'invented' because in times gone by it was hard to get power devices with decent hfe values. Nowadays these integrated Darlington devices seem less common but I have a great deal of respect for most all of what Sanken makes, their transistors are often praised for their quality.
I'm not sure how much cost benefit was obtained by the use of power Darlingtons in my Pioneer amplifier, they could have used cheaper parts. It doesn't even use SMT transistors at the front end which would be cheaper too. I've opened up really cheap amplifiers and they didn't use power Darlingtons. Pioneer have their reasons but I"m not sure what they are.
Certainly, there is more flexibility when you build a discrete Darlington and there are some really superb hi-end devices available, especially the older Japanese parts. I always assumed that integrated Darlington devices were 'invented' because in times gone by it was hard to get power devices with decent hfe values. Nowadays these integrated Darlington devices seem less common but I have a great deal of respect for most all of what Sanken makes, their transistors are often praised for their quality.
Last edited:
They are not. However, notice the exponential (logarithmic) nature of the current gain. This is an undesirable transfer and not being able to get to the driver device means it is what it is. This is the reason Dr Hawksford came up with his 'amplified diode' feedback/feedforward error correction circuit. It just so happens that vertical type mosfets have a similar transfer and so the amplified diode circuit can be applied to them with some minor adjustments to create an overall linear transfer from the output stage. Don't be misled by the thinking that global feedback fixes all that non-linearity, the bandwidth required to do so is way out of range for a global loop spanning multiple stages and PCB traces. The distortions get pushed into a higher band that is usually filtered out by Miller or other types of compensation. Those folks who tend to say that feedback 'fixes' everything generally swamp the global loop with excessive compensation in order to gain stability, limiting bandwidth in the process. You must strive for linearity of each stage and the interaction of the cascade stages before global feedback is applied.😉🙂
This paper is from 1981, so nothing really new here...🙄
This paper is from 1981, so nothing really new here...🙄
Last edited:
I don't see exponential in the hfe vs IC. But I do notice the hfe rise up first upon increase current and then drop. The curve looks like a big hump. I looked and a typical power BJT, the hfe remains flat until high current before it starts dropping. Is that what you are referring to. That really makes sense. So no darlington!!!
Is this also the reason not to use darlington on the pre-driver?
Thanks for the article.
Last edited:
The graph is logarithmic in scale. @Ic=200mA, Hfe is ~2600, Ib=77uA. @Ic=1A, Hfe is 8500, Ib=118. @Ic=4A, Hfe=15,000 Ib=267uA. @Ic=7A Hfe=12,000 Ib=580uA. @Ic=10A Hfe=10,000 Ib=1mA. @Ic=12A, Hfe=8500, Ib=1.4mA.
Although the current gain is very high, it does not seem like a very linear d(t)Ic/Ib to me.
Generally, bipolars for linear audio tend to have a much more linear current gain up until the point of electron crowding, indicated by the beta droop as Ic increases.
Darlingtons are fine if they are used with the proper driving circuitry as are vertical mosfets. The nasty looking transfer and folks who use them as if they are linear bipolars and thus get poor results is what gives them a bad reputation.🙂 Darlington drivers are generally not necessary because small signal transistors tend to have a much higher current gain and can easily be accompanied by a small signal pre-driver stage.
Although the current gain is very high, it does not seem like a very linear d(t)Ic/Ib to me.
Generally, bipolars for linear audio tend to have a much more linear current gain up until the point of electron crowding, indicated by the beta droop as Ic increases.Darlingtons are fine if they are used with the proper driving circuitry as are vertical mosfets. The nasty looking transfer and folks who use them as if they are linear bipolars and thus get poor results is what gives them a bad reputation.🙂 Darlington drivers are generally not necessary because small signal transistors tend to have a much higher current gain and can easily be accompanied by a small signal pre-driver stage.
Last edited:
Ha ha, I missed that!!!! Now I am convinced that darlington is no good. I am not going to do the compensation, just use simple BJT. I knew there is a good reason for people not to use it, ha ha, at least I knew better to run this by you guys first instead of being a smart alec.
I am not going to add any distortion compensation if I can help it. In Chinese, we have a saying, the more shrine you put, the more ghost you attract!!!😱😀😛 That's the reason I decided against MOSFET as they need distortion compensation also.
Thanks for clarifying all these. Now I feel better.
Last edited:
A monolithic Darlington eliminates SOME parasitic effects when compared to a discrete arrangement. Got to keep them parasites at bay.
Side question. Is 3EF really improve that sound? It is troublesome!!! A lot of high end amps don't use 3EF. I know YBA that got me into all these don't use 3EF.
I'll spell out my argument in more detail.
An old fashioned quasi comp amp has the positive side of its output stage Darlington and the negative side Sziklai pair.
The Baxandall diode alters the transfer function of the negative side to match the positive side better.
Therefore (I argue) by adding Baxandall diodes to both sides of a modern CFP output stage, it would get the broad crossover region of a double EF, but keep the thermal stability of the CFP, as the diodes hardly dissipate any power.
Keep in mind that Hfe is referring to DC beta. AC beta is frequency dependent, as every transistor has junction capacitance that drags it down and must be considered in the AC transfer figure. THD-20KHz will be larger than THD-1KHz. It may be better to have 3EF, or maybe not.
- Status
- Not open for further replies.