Not the copper traces for audio , but the semi's. Compensation is just
manipulation of this delay. Too much delay = an oscillator.
For a PC MB , the traces themselves DO matter . (wavy traces at 1Ghz).
OS
The really serious problem in bipolars is storage time when the device saturates. That will assure cross conduction for sufficiently high frequencies. So one needs to be certain that the output devices don't get too low of a |Vce|.
DMOS does not suffer from storage time, but they do have large capacitances to drive.
DMOS does not suffer from storage time, but they do have large capacitances to drive.
Someone queried comments I made in an earlier post, about time delays due to stored charges being a cause of phase shifts.
I sensed an air of surprise in the question and I looked for some transistor datasheets to try to quantify switching times such as storage time or ts, to make my argument more convincing.
All I could come up with was some information on high voltage switching transistors tested with a current of 1 ampere which is not direct enough as evidence.
Are you able to help with the relevant sort of information?
A standard treatment is in Sze and Ng, Physics of Semiconductor Devices, 3rd edition, pp. 271-273 (ISBN 978-0-471-14323-9. At the risk of copyright violation and spine damage I'll try to scan it and post.
The comments about shortening on/off times came from post 18236, while I was mulling over the content for my post 18237 containing information about Carrier storage taken from an electronic dictionary source.
I doubt anyone reading this would want to drive the bases harder.
The members argument seems to be more a case of running the output stage in Class A at low levels with rare excursions into the Class B realm. I have seen this consistent line of reasoning often enough in previous contributions of his.
Last edited:
Thanks for quoting the source and offering to scan. If you could be so good as to give some figures for a typical audio grade transistor that would do fine.
I have a few old spine damaged references too.
Yes running rich class AB is fine, although there is a gain variation for the standard complementary emitter follower as a function of input/output voltage. This changes loop gain and potentially affects stability margin if you are on the edge already. There are good discussions in Cordell's power amplifier book I seem to recall.
But this smooth modulation is infinitely better than dead zone stuff from underbiased class B.
That may be difficult to find, as most of the audio-preferred devices are not simultaneously characterized for switching.
Here's one that is characterized for switching and power amplifiers. Note the misprints of "m" for micro at the beginning of the datasheet.
http://www.onsemi.com/pub_link/Collateral/2N6338-D.PDF
http://www.onsemi.com/pub_link/Collateral/2N6338-D.PDF
I generally can if I bother to try, and it's always in the bass lines, especially if the programme contains a lot of fast bass line attack.
Mjona, Bcarso
Your views are duly noted, but gentlemen, there are many Ifs in between almost pure class B and a well biased class AB. This depends on a number of factors, one if which is how well can you drive the output stage in general, which power devices are you using and is what you are using up to the design goals, etc. As an example, it's hardly all the same if you want to drive two pairs of trannies at 200W/4 Ohms, or is it 4 pairs. The regime those devices will have to operate under is quite different if more share the load.
Also, biasing has to be reasonable. I usually look for at least 1W/4 Ohms in pure class A, and that's usually enough, more could be bad, but is not necessarily so. On the other hand, I am VERY exclusive in selecting my output devices and by default stick to ON Semi's NJL or NJW series because I know the rather well. I would not be so liberal if I was somehow forced to use Sanken devices, which I dislike and hence don't really know too well.
Your views are duly noted, but gentlemen, there are many Ifs in between almost pure class B and a well biased class AB. This depends on a number of factors, one if which is how well can you drive the output stage in general, which power devices are you using and is what you are using up to the design goals, etc. As an example, it's hardly all the same if you want to drive two pairs of trannies at 200W/4 Ohms, or is it 4 pairs. The regime those devices will have to operate under is quite different if more share the load.
Also, biasing has to be reasonable. I usually look for at least 1W/4 Ohms in pure class A, and that's usually enough, more could be bad, but is not necessarily so. On the other hand, I am VERY exclusive in selecting my output devices and by default stick to ON Semi's NJL or NJW series because I know the rather well. I would not be so liberal if I was somehow forced to use Sanken devices, which I dislike and hence don't really know too well.
Hi Dvv.
It might be VLF 1/f noise that you are objecting to.
IME, this vlf noise destroys 'solidness' in the bass.
Try playing pink noise through your system and listen closely.
It is as though the pink noise source excites 1/f noise behaviour in the system....excess noise.
LF pink noise should sound low and solid.....1/f noisy system playback will cause a rumbly/bursting/memory character that does not belong.
This VLF 1/f behaviour has consequences throughout the audio spectrum also.
Dan.
Last edited:
At last we start to see our way out of the forest. Each amp has it's ideal limiting frequency ( LF also ). To be honest if the amp can give 30 kHz - 1 dB I would say job done ( 5 Hz - 3 dB or better ). Thus even the humble 2N3055 can make beautiful sounding amps. Douglas Self's books shows you how to make the AK47 of amps. One that will work and often puts " better " amps to shame. The problem being those amps are so complicated how they exactly work may never be known. The medical world is no better. Yesterday I learned they have no real idea how alcohol works in our bodies. They just have a very fair idea . I am not kidding. Taking the maximum recomended weekly dose which no one I know sticks to ( except me because I am so mean ) all have liver damage. It is 21 units or I guess 3 x 75 mL of wine in 7 days. 5 litres of beer perhaps if the lager type. The crazy thing is if your liver is divided in two it will grow back in the same time as to repair liver damage by following guidlines on wine. I kid you not.
Good point, Dan, you may well be quite right. Frankly, since I am not in the least interested in output MOSFETs, I never bothered to investigate for the reason. Your suggested test seems straightforward anugh, all I need now is the time to do it. Time always seems to the major problem, you know what I mean?
IRF240/9240 have one virtue. They are cheap. Exicon 10N/P20 in quads should do any job. Ron will be slightly higher than one set of Bipolar if with 0R33.
Most speakers have an inductor in series with the bass unit if passive. If someone states bipolar have more bass in this situation I doubt it.
The great thing about the true audio MOSFET's is bias is set by ears and fingers. Anything else is not really required. If the bias is high the bass will suffer. If the bias is zero it is almost possible ( 3 mA usually ). At about 1V between gates bias starts to be about right. The IRF devices need a Vbe bias device like bipolars and far more bias voltage. There are if you like 3 rd worse choice.
Most speakers have an inductor in series with the bass unit if passive. If someone states bipolar have more bass in this situation I doubt it.
The great thing about the true audio MOSFET's is bias is set by ears and fingers. Anything else is not really required. If the bias is high the bass will suffer. If the bias is zero it is almost possible ( 3 mA usually ). At about 1V between gates bias starts to be about right. The IRF devices need a Vbe bias device like bipolars and far more bias voltage. There are if you like 3 rd worse choice.
Nige, in a slow progression, I managed to both simlify my power amp and to push it upwards where I never thought I'd get to. With input low pass filter excluded, the 1W frequency response is now around 1.6 MHz. Full power bandwidth is a bit above 1 MHz, far too wide even for me, so there's an input low pass filter to control that.
All achieved with off-the-shelf parts, with a reasonable cost and well within the reach of a DIYer with at least some experience. Output stage uses 4 NJW pairs of devices, and it's biased at 130 mA per device I'll be sending it to Alex within a few days. Global NFB 20 dB, THD and IM at <0.03% into 4 Ohms.
All achieved with off-the-shelf parts, with a reasonable cost and well within the reach of a DIYer with at least some experience. Output stage uses 4 NJW pairs of devices, and it's biased at 130 mA per device I'll be sending it to Alex within a few days. Global NFB 20 dB, THD and IM at <0.03% into 4 Ohms.
- Status
- Not open for further replies.