Biasing Question

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I would like to replace the output mosfets IRF640 and IRF9640 with Toshiba 2SK1530 and 2SJ201. After looking at the data sheets I noticed that Vgs on the Toshiba has a lower voltag.

Vgs Irf640 is from 2v to 4V

Vgs 2sk1530 is from .8v to2.8V

Seems to me that the Bias point will have to be adjusted but I am not sure which resistors need to be changed or if anything else should be changed. If anyone has any ideas please let me know.

Thanks
 

Attachments

  • mosamp.jpg
    mosamp.jpg
    130.2 KB · Views: 306
I foretell a blow up coming.
:Pumpkin: :D
quite right.

Reducing R5 & R23 would reset the DC bias but there is a problem. It also reduces the gain of that stage (T1 & Q8) to less than 1. This means the Op-Amp must output a larger voltage swing. In addition, comparing the transfer functions, 2sk1530/2sj201 requires a larger swing in Vgs for the same conductance, requiring more voltage swing from the Op-Amp.

Thermal connectivity should be between the bias spreader and the driver transistors (T1 & Q8), as is the scheme for CFP OPS. I do not see a requirement for the IRF510 in the circuit. A typical Vbe multiplier should work. This allows you to reduce the bias voltage across R17/R24, reducing the value and lowering the AC impedance at the emitters, increasing gain. The ouput transistors should have source ballast resistors added so there is more equal current sharing and to help 'normalize' the transconducance through the current crossover.
 
I used this same circuit in a car amplifier that I built a long time age. if I remember correctly, I had to match all of the output devices in order to have equal current sharing. I had considered adding the ballast resistors simply because it is very $$ to buy a bunch of mosfets just to find out that they don't match. I would think that ballast resistors would be around .1ohm I am new to the forum and am not familure with the Schematic you mentioned CFP OPS.

Thanks
 
Current Feedback Pair. CFP provides voltage gain at the output stage. Current is fed back to the emitters of the drivers establishing a set loop gain for the output stage. In this type of circuit it is the drivers that need to be monitored with a Vbe multiplier for thermal runaway not the output devices. But there would need to be a few changes in the circuit to use 2sk1530/2sj201.
 
I'm re-biasing a pair of monoblocks.

Originally they had approx 10mV voltage drop accross the 0,5R emitter resistor, which translates to 20mA bias current per transistor. There are total 24 transistors in this amplifier. As it is a bridged balanced amp, the other amplifier board has 12 and the other 12 transistors. The biasing needs of course to be done on both boards.

The heat sinks support approximately 45mV bias voltage (90mA current). This rises the heat sink's temperature to 47 degrees (room temperature 27 degrees, the hottest time of the year), the final temp might be little higher once I place the amps on equipment rack. The distortiuon figeres, which I measured with E-mu 0404 USB and ARTA, looks promising; H2 drops significantly (~20dB), but H3 rises somewhat (0,0025% -> 0,0045% @ 10kHz = 6dB increase). Why the H3 rises albeit the bias is increased and the transistor should work more linearly and stable? The distortion spectrum in H3-dominated as the biasing reduces the H2 so much.

How do I calculate the actual A-class region of such an amp, does the fact being balanced amp change something? 90mA times 12 (NPNs or PNPs in the amp) = 1080mA (or should I calculate with all 24 transistors = 2160 mA?). How much power does this amp produce in A-class with such bias? (9,3w/18,6w or something else?)

All help and comments much appreciated

Regards,

Legis
 
Last edited:
If the output bias is 1080mA, i.e. 1080 bias flowing throw the 12 NPNs and the same 1080mA flowing through the 12 PNPs, then the ClassA current limit of this Push-Pull stage is 2160mA.

The maximum ClassA output power is Ipk^2 * Rload / 2 = 2.16*2.16*8/2 = 18W

10mVre is unusually low for an EF output stage. Is your amp fitted with a CFP output stage?
But, equally, 45mVre is well above an optimal ClassAB bias voltage.

Q.
does the bridged balanced amp have 24 output transistors or 48 output transistors?
If 24, then it must have a 6pair output stage in each balanced half and the 10mVre total bias current is only 120mA for a ClassA current limit of 240mA giving 1/4W of ClassA output.
 
Last edited:
If the output bias is 1080mA, i.e. 1080 bias flowing throw the 12 NPNs and the same 1080mA flowing through the 12 PNPs, then the ClassA current limit of this Push-Pull stage is 2160mA.

The maximum ClassA output power is Ipk^2 * Rload / 2 = 2.16*2.16*8/2 = 18W

10mVre is unusually low for an EF output stage. Is your amp fitted with a CFP output stage?
But, equally, 45mVre is well above an optimal ClassAB bias voltage.


Hi Andrew,

I could use some common English :D. I have no idea, what are the charachteristics/differences of EF and CFP output stages? Here you can see the one amp PCB of the amplifier: http://www.diyaudio.com/forums/solid-state/171491-emotiva-xpa-1-vs-jungson-99d-12.html#post2667764

Yep, I read something about commonly recommended optimal bias voltage of 22-23mV. Tried that, did not notice anything special, the H3 had already risen so upping the bias beyond this comes "free". I would quess somewhere around 14mV does not change the H3, or if it does, it is very slight.

I also noticed that the amp distorts little more when driven past the A-class region. With low bias the behaviour at high leves is better. This is common trade-off between high/low bias I quess?
 
That looks like a 6pair output stage.
bias to 22mVre for a total bias of 264mA.
The maximum ClassA is ~ 1W

Yep, that is the other of the two identical amp boards inside one monoblock, total 12 pairs of BJTs.

Why the 22mV bias voltage, which I have also heard, is good? What makes it good, or better than higher bias with AB-class amps? THD is significantly lower with 45mV bias.
 
Can you see/measure the crossover distortion (CD)?
You may have to measure the CD when the output is 100mW to 1W, not at quarter power and certainly not at near full power.

At that low levels all harmonics besides H2 and H3 are below the noise floor in FFT screen of ARTA so I cannot see them. I have an oscilloscope but I doubt it could pick up something like that, and there's also that I don't know exaclty how to measure this with scope.
 
Crossover distortion does not contain any significant H2, nor H3.

It's more likely to contain H9, H11, H13, etc.

How does one measure XO distortion the easiest?

If the bias is unoptimal, what region (<A-class zone, middle, full power) gets affected primarily? Is the XO distortion greater even in "A-class region" or is it just bigger after that region, compared to optimally biased amp?
 
This book suggests on page 74 that 26mV (not 22-23mV) is the theoretical optimal bias voltage for minimizing XO-distortion. I decided to bias the amps with this value, as it also gives reasonably high bias current (~1,25A) without a need to change the emitter resistors to lower value.
 
26mVre is the total voltage across the external RE and the intrinsic emitter resistance (re?).

That internal resistance is quite low, particularly at the output bias currents that are adopted.
As a result a bigger proportion of the resistance and thus the Vre appears external to the output transistor.
With very low Re, say ~0r1 then the minimum crossover distortion (optimal bias ClassAB) may occur at 15mVre to 20mVre.
A higher Re, say ~1r0, then the optimal bias may be 24mVre to 26mVre.

I believe that this optimal bias changes with temperature (absolute temperature).
Tj = 50degC is an increase of 25Cdegrees above 273.16K + 25C or 323.16K (an increase of ~18% in junction temperature ref 25egC). The optimal bias for Tj=323.16K is somewhat higher than for 298.16K

A distortion meter measures, to it's own limits, all the distortions.
Cordell, JLH and others tell you how to build one.
Basically very accurately notch out the fundamental while preserving all the harmonics from H2 to Hxxx and measure the residual as a percentage. Viewing the residual on a scope along with an attenuated fundamental is very instructive. There are many scope examples on the web.
 
I have HP 54503A scope, that has Low frequency reject mode, that has -3dB point in 450Hz. Could I maybe watch a residual for example 50-100Hz sine with this scope without hi-pass filter? I'd like also to measure at 10-20kHz as the CD is strongest in that region. Do they sell steep (min. 18dB/oct.) hi-pass filters for scopes somewhere?

Could I also look at the "difference" (don't know correct term) between the waveforms? If I would take the original signal of ~1V from soundcard to channel 1 and the amplified signal of ~1-20V from the amp to channel 2, and compared them on scope, would I get something uselful out this way? Or should the voltage of the signals be identical for this to work?
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.