bipolar (BJT) transistor families for audio power output stages

I have no problems with oscilations, the faster device gives me better performance though, probably because happen to have very low cob . With very low cob transistors I find stabilty is much easier to achieve.

If stability was such a major problem no one would be using vertical mosfets either.

Reliability with slower Onsemi is better................., why not reliability with slower japanese devices, someone must be getting a lot of freebies and samples, pity one cannot get this from the japanese.

Output stage parasitic oscillations are merely a circuit design/layout concern that is made more challenging by fast devices. In general, we want fast devices and just need to put up with the greater attention to detail they need. It is indeed true that vertical MOSFETs need more care and feeding in this regard, with their equivalent ft values in the hundreds of MHz.

A base-collector or gate-drain Zobel network can be very helpful in this regard because it kills the Q of any resonance involving the base-collector capacitance or gate-drain capacitance. If you study up on Hartley and Colpitts oscillators you will see that with a little bit of stray inductance you can usually form one with an output transistor. This was discussed briefly in my JAES MOSFET amplifier paper.

Cheers,
Bob
 
What I am trying to say is, that fT about 4-7 MHz is more than appropriate for very well performing amplifier for audio, no need to use output devices with fT 50Mhz.. . Problems with speed and linearity are very often in stages before output stage. How would You explain many amplifiers, using best output devices, but at the same time with poor linearity , with a lot of IMD products and stability problems wiht complex load? And mostly very expensive amplifiers, and often class A..
As You said, design is main problem (e.g. switch off problem..), not output devices with 5 or 50MHz fT .
And samples.. I had no one , all (several thousands) payed. Reliability means also You can use 2 pairs instead of 3 or 4 pairs, so in final lower costs and lower counts means better reliability (e. g possible thermal pad failure).

I disagree. High ft output transistors are very desirable and ft of 4-8 MHz will yield a good, but not the best, amplifier.

Assume that a good front end and VAS is a given. The output stage is where it is difficult to get really low HF distortion in class AB. Higher ft devices mean lower HF distortion, all else remaining equal. It is important to understand dynamic crossover distortion (sometimes called secondary crossover distortion) and shoot-through conduction in this regard.

High ft output devices do two things for you:

1) they enable higher output current rate of change, leading to less dynamic crossover distortion.

2) they result in less excess phase shift in the output stage, allowing greater feedback stability and/or a higher feedback gain crossover frequency.

Cheers,
Bob
 
i appreciate your patience ,bob, and we re in the process to be fully agree..
what you said is right , and i will add that it s not only the Ft, but
the ability of the device to sustain this Ft at the higher frequency
possible...

what i tried to explain you with my flawed english is that the Ft as
defined by datasheets, is in fact frequency dependent..
for a given current, the bjts have not the same Ft at 1 MHZ and at
5 MHZ....this is where the japanese parts are better than onsemi s :
they sustain their Ft in a wider frequency range....


regards,

wahb

I may be wrong here and may learn something, but it is my understanding that ft is not generally a function of frequency. At a given operating current, you have a given ft for a given device. That ft is reflective of the fact that the ac beta is declining at 6 dB/octave at frequencies above f_beta.

If you have ft of 20 MHz, beta will be 20 at 1 MHz and 5 at 4 MHz and approximately unity at 20 MHz (barring a large influence of internal transistor parasitics like base resistance).

When you run SPICE it reports ft of the device at the operating point; it does not care about that as a function of frequency.

Where did you see a test circuit from OnSemi that was used to measure ft? It is not in the datasheet.

My understanding of measuring ft does not include having a low-value resistor between the signal source and the base. Are you confusing this with some kind of switching time test circuit?

My understanding of measuring ft is that you measure it by measuring ac beta at some high frequency (like 1 MHz), then inferring ft from that. ft in MHz equals ac beta at 1 MHz. AC beta is measured with a known moderate resistance in the base line (like maybe 1K), or with a high-frequency AC current source, then looking at AC collector current.

You can easily simulate this test setup in SPICE and see how the AC current gain declines with frequency above a certain frequency (f_beta).

Maybe Andy_C can chime in here.

Cheers,
Bob
 
I may be wrong here and may learn something, but it is my understanding that ft is not generally a function of frequency. At a given operating current, you have a given ft for a given device. That ft is reflective of the fact that the ac beta is declining at 6 dB/octave at frequencies above f_beta.

If you have ft of 20 MHz, beta will be 20 at 1 MHz and 5 at 4 MHz and approximately unity at 20 MHz (barring a large influence of internal transistor parasitics like base resistance).

When you run SPICE it reports ft of the device at the operating point; it does not care about that as a function of frequency.

Where did you see a test circuit from OnSemi that was used to measure ft? It is not in the datasheet.

My understanding of measuring ft does not include having a low-value resistor between the signal source and the base. Are you confusing this with some kind of switching time test circuit?

My understanding of measuring ft is that you measure it by measuring ac beta at some high frequency (like 1 MHz), then inferring ft from that. ft in MHz equals ac beta at 1 MHz. AC beta is measured with a known moderate resistance in the base line (like maybe 1K), or with a high-frequency AC current source, then looking at AC collector current.

You can easily simulate this test setup in SPICE and see how the AC current gain declines with frequency above a certain frequency (f_beta).

Maybe Andy_C can chime in here.

Cheers,
Bob


as i explained in a earlier post, you can t transpose the aop reasoning scheme.
i already described how the Ft as defined by datasheets is not representative of the current gain versus frequency...infering is not possible..
you cant exchange current beta for frequency, like in the op amps gbw product..

i gave the exemple of the MJL that have a 50 mhz Ft at 1A and 1MHZ..
that make a beta of 50....

if you change the frequency to 10 mhz, the current being the same,
the fT WILL BE MUCH LOWER, and not be 50mhz as expected, as the
gain will be lower than the 5 that would be logically calculated..


the MJL and 2SC5200 have comparable Ft...
here a graph of bandwith for a constant current of 4A..
we can see that the japanese part sustain his Ft at much higher
frequency that the onsemi part..
the AC beta decline more rapidly for onsemi s...

the test is in common emitter mode, but i can assure you thatit s the same in emitter follower mode...

regards,

wahab
 

Attachments

  • CE BW.zip
    5.5 KB · Views: 72
What I am trying to say is, that fT about 4-7 MHz is more than appropriate for very well performing amplifier for audio, no need to use output devices with fT 50Mhz.

i already described how the Ft as defined by datasheets is not representative of the current gain versus frequency...infering is not possible..
you cant exchange current beta for frequency, like in the op amps gbw product..

I may be wrong here and may learn something, but it is my understanding that ft is not generally a function of frequency.

I disagree. High ft output transistors are very desirable and ft of 4-8 MHz will yield a good, but not the best, amplifier.

You are correct Bob, and me too :) BV, wahab, perhaps it will help recalling the following:

- Fc, or the cut-off frequency, is defined as the frequency at which the real and the imaginary part of the common emitter input impedance are equal. In a first approximation, this means that 1/jwCb'e = Beta/gm, hence the Beta dependency on frequency and collector current (through both gm and Cb'e).

- Fc defines a first order system, with a pole at Fc, where beta drops 3dB from the LF value.

- Extrapolating the Beta frequency dependency yields the transition frequency, defined as the frequency at which Beta = 1. As for every first order (single pole) system, this dependency is the gain x bandwidth product.

- A Ft=5MHz and a Beta at Fc of 100 yields a cut off frequency of only 50KHz. Otherwise said, the power device quadripole equivalent h21 (gain) parameter has a pole at 50KHz.

- Now look at an emitter follower as a (local) negative feedback stage. The base amp gain has a pole at 50KHz and feedback, setting the close loop gain to unity, also sets stage bandwidth: 50KHz.

- Now, unless some very special methods are applied, the entire open loop amp (input stage, VAS, output follower) will have a pole at the same 50KHz. It would be very difficult to provide enough open loop gain to reduce the HF (20KHz) distortions.

Bottom line, a Ft = 5MHz is not good enough for a high performance circuit.

- Consider now a device with Ft=50MHz and repeat the above; we are now talking about Fc of 500KHz and everything above looks much better.

This is an oversimplified small signal analysis. There's nothing above dealing with the large signal effect, like crossover conduction in a class AB stage, whic is another kettle of fish.

BTW, Bob is also correct about measuring Ft @ 1MHz. The idea is to measure the beta drop at 1MHz (supposingly > than Fc above), then extrapolate to beta=1 (using the measured value as a gain-bandwidth product) to extract Ft. For all practical purposes, the gain-bandwidth approximation is very good for medium frequency devices (say, for Ft < 300MHz). Higher frequency and microwave devices are characterized in completely different ways.

P.S. I guess part of the confusion is coming from neglecting the fact that Ft has a peak depending on the collector current. It only makes sense to compare apples to apples when dealing with different devices, that is, estimating the Ft at the peak Ft current.
 
Last edited:
Bob,
the importance of speed cannot be underestimated, the lack of speed is the underlying cause of all distortions, through various mechanisms, creating by nature always harmonically unrelated transients. Slewing is caused by internal capacitances, in absence of them the bandwidth would be infinite, but it`s also about electron mobility. Solid states devices are slow, responding slowly to a change, bringing about phase shifts and a devastating time difference between input and output. That´s why they distort so disagreeably. The ear does not bother with amplitude distortion, but cannot stand time errors.
 
Bob,
the importance of speed cannot be underestimated, the lack of speed is the underlying cause of all distortions, through various mechanisms, creating by nature always harmonically unrelated transients. Slewing is caused by internal capacitances, in absence of them the bandwidth would be infinite, but it`s also about electron mobility. Solid states devices are slow, responding slowly to a change, bringing about phase shifts and a devastating time difference between input and output. That´s why they distort so disagreeably. The ear does not bother with amplitude distortion, but cannot stand time errors.

Spambot nonsense, as usual.
 
  • Like
Reactions: 1 user
syn08, you miss the fact that Ft are measured with very low impedance
sources, thus rendering the influence of Cbe irrelevant for a large
part of the curve..
unity gain as measured by sims is for a 4A current and 5V VCE , 2R impedance source :
MJL3281 : 6MHZ....
2SC3281 : 10 MHZ..
2SC5200 : 120 MHZ..

apart from the last bjt, which is frankly optimistic,
how do yopu explain those differences, since the MJL3281 has
on datasheets better Ft than the 2SC3281, and on par with the
2SC5200...
does onsemi plague his models?..or is it toshiba that release
false models?...
have you some clue about this strange behaviors..?..
your help would be greatly appreciated removing these paradoxes..
 
syn08, you miss the fact that Ft are measured with very low impedance
sources, thus rendering the influence of Cbe irrelevant for a large
part of the curve..
unity gain as measured by sims is for a 4A current and 5V VCE , 2R impedance source :
MJL3281 : 6MHZ....
2SC3281 : 10 MHZ..
2SC5200 : 120 MHZ..

apart from the last bjt, which is frankly optimistic,
how do yopu explain those differences, since the MJL3281 has
on datasheets better Ft than the 2SC3281, and on par with the
2SC5200...
does onsemi plague his models?..or is it toshiba that release
false models?...
have you some clue about this strange behaviors..?..
your help would be greatly appreciated removing these paradoxes..

Wahab,

I think you need to show the test circuit you used to measure/simulate ft, and used for the plot in your earlier post. You keep saying that ft is measured with very low impedance. Please show us what you mean with a schematic and let us know where you got this. This 2-ohm impedance source is not making sense to me, but I may just be slow on the up-take. I think if you show the test circuit you are using it will be more clear where that 2-ohms is, and we will be better able to see for ourselves what its influence is.

Cheers,
Bob
 
Wahab,

I think you need to show the test circuit you used to measure/simulate ft, and used for the plot in your earlier post. You keep saying that ft is measured with very low impedance. Please show us what you mean with a schematic and let us know where you got this. This 2-ohm impedance source is not making sense to me, but I may just be slow on the up-take. I think if you show the test circuit you are using it will be more clear where that 2-ohms is, and we will be better able to see for ourselves what its influence is.

Cheers,
Bob

here the configuration...
if the 2R resistance is increased, Ft collapse dramatically..
each current source is adjusted so the Ic is 4 A...
 

Attachments

  • CE CONFIGURATION.zip
    7.6 KB · Views: 132
Gentlemen, I know very well what are You talking about, and what I am talking about..We never used power BJT´s in isolation, and final result depends much more on whole amp desing that on fT of power devices used(in some range, of course). As I posted earlier in post 14 ( SET) , this is quite simple amp using 2 pairs of MJL21193/21194, speed and removing base charges are not an issue.. Show please some practical results and similar measurements with Your´s prefered BJT devices.