Currently I am running ~300mA of bias through the output stage. The heat sink temperature is ~55C. The ambient temp is ~35C. If the ambient is reduced to ~27C, what will I be able to increase the bias level to, bringing the heatsinks up to ~55C again?
My best guess is about ~380mA's
Idea's??
Thanks, DonS
My best guess is about ~380mA's
Idea's??
Thanks, DonS
Don S said:
You could try returning the 1k feedback resistor to ground via a big electrolytic. That will give a DC gain of unity, so the offest voltage of the MOSFET input stage will be amplified by much less.
Cheers,
Glen
G.Kleinschmidt said:
Thanks Glen. I have been considering that idea. I was hoping to find another solution without changing the LF performance. Also the idea of an electrolytic in the signal path is unappealing to me on principal. I was actually thinking of removing the lytic across the VBE multiplier!
Maybe since it is a negative offset I could run a little more current through the diff pair and see if that improves things.
Idea's and suggestions are welcome!
Thanks, Don
Don,
Glen's solution is routinely used, entirely conventional and works well. The cap is in shunt, and its effect sonically is much less than if it were in series with the signal. Try it, it can give outstanding results.
Cheers,
Hugh
Glen's solution is routinely used, entirely conventional and works well. The cap is in shunt, and its effect sonically is much less than if it were in series with the signal. Try it, it can give outstanding results.
Cheers,
Hugh
Hi Don,
I thought I would lend a late comment here.
The 0.1 R resistors were due to the normally tight matching on Japanese devices. The new On Semi parts are also pretty good. You don't have to buy the normal tonnage to get well matched sets (as Hugh is use to doing! 😉 ). My suggestion would be to retain the original emitter resistor values as to go higher may affect the sound quality. Did it use those flat non-inductive parts? Don't use normal 5W "bathtub" style resistors. They are inductively wound.
I have more experience with MJ21195 and MJ21196. They are very tolerant of problems, but you can blow them. If the amp lasted as long as it did, just be happy and rebuild it. The MJ21193 and MJ21194 are very closely related.
I fully agree with Andrew and the other posters with regard to output pairs and load values for a new design. However I do recognize that you only have this to work with. I think you will be fine. Just don't run "silly speakers". The Kappa series, Ohms, stuff like that. You might be further ahead if you increase the voltage rating on your supply caps. The capacitance is not such a high concern. Your amp will only draw what it is asked for. The compression on the odd peak pulse is what works to save the amp.
For what it's worth Don ......
-Chris
I thought I would lend a late comment here.
The 0.1 R resistors were due to the normally tight matching on Japanese devices. The new On Semi parts are also pretty good. You don't have to buy the normal tonnage to get well matched sets (as Hugh is use to doing! 😉 ). My suggestion would be to retain the original emitter resistor values as to go higher may affect the sound quality. Did it use those flat non-inductive parts? Don't use normal 5W "bathtub" style resistors. They are inductively wound.
I have more experience with MJ21195 and MJ21196. They are very tolerant of problems, but you can blow them. If the amp lasted as long as it did, just be happy and rebuild it. The MJ21193 and MJ21194 are very closely related.
I fully agree with Andrew and the other posters with regard to output pairs and load values for a new design. However I do recognize that you only have this to work with. I think you will be fine. Just don't run "silly speakers". The Kappa series, Ohms, stuff like that. You might be further ahead if you increase the voltage rating on your supply caps. The capacitance is not such a high concern. Your amp will only draw what it is asked for. The compression on the odd peak pulse is what works to save the amp.
For what it's worth Don ......
-Chris
AKSA said:
Thanks, Glen and Hugh. I will try it and see how it works.
I have 2 questions though.
If I use a 330uF value is the F3 0.5HZ? Is my math correct?
How should I set the polarity of the cap? Positive or negative to ground? I have an initial negative offset, so my instincts tell me to put positive to ground. ???
Thanks, Don
Don,
Your gain is set to 11, which seems a tad low to me. You can increase the gain to around 20 (26dB) by simply reducing the shunt resistor (R12?) from 1K to 500R.
Where there is no bias current with mosfets you need not ensure that the impedances on input and feedback node are the same.
A 500R shunt resistor would have a 6dB corner with a 33uF cap at 10Hz. To improve radically on this, 100uF would be fine. That would give you 6dB at around 3.3Hz, which is closer to the mark. You really don't need to go lower than this; to my knowledge there is no musical recording around with signals below about 25Hz.
Cheers,
Hugh
Your gain is set to 11, which seems a tad low to me. You can increase the gain to around 20 (26dB) by simply reducing the shunt resistor (R12?) from 1K to 500R.
Where there is no bias current with mosfets you need not ensure that the impedances on input and feedback node are the same.
A 500R shunt resistor would have a 6dB corner with a 33uF cap at 10Hz. To improve radically on this, 100uF would be fine. That would give you 6dB at around 3.3Hz, which is closer to the mark. You really don't need to go lower than this; to my knowledge there is no musical recording around with signals below about 25Hz.
Cheers,
Hugh
anatech said:
Thanks, for replying Chris!
I have settled on 0R235 for the emitter resistors. The originals were little flat packs like you thought. The new ones are metal oxide devices. I actually use (2) 2watt'rs in parallel to get the value. My concerns were bias stability and current hogging. The original amp used a very complex circuit for the bias control. Dual opposite polarity darlingtons. (One is near extinct.) They also very tightly matched the outputs in the circuit! (I went for KISS.) It seems to work very well and sounds good too!
The original circuit board's are toast! They have delaminated just about everywhere. For me it was easier to design a new amp than to recreate the old one. Now the output stage is on a separate board from the FE and I can change the FE board easily if I see fit.
The one thing that I guess I didn't get across well is that 3 pairs of MJ21193,4's should have a larger SOAR than the Sanken's removed! Not that they can drive a dead short.
If I were doing a ground up design I would use ~60V rails for the outputs and use at least 15 pairs for each channel. 100,000uF of rail capacitance per channel minimum. ~70V rails for the FE with active filtering. As large heatsinks as I could afford, and bias it as close to class A as possible into 8R.
Thanks, Don
AKSA said:
Hugh, I have lowered the C/L gain of the circuit on the suggestion of one very respected designer on this site. This was in regard to the FE. When I was running it at about 30DB C/L gain.
"On the other
hand, you are operating near open loop - your feedback loop
is asking for almost 40 dB of gain and getting 30." "Keep in mind that with the current circuit you would probably
not want to operate it at greater than 20 dB or so."
I have made a few small changes since then, but I don't know how to calculate O/L gain. I am in the low FB camp. However the amp does seem to sound better in most respects at 20DB C/L gain. A couple of things might have been better at 26DB though. The best compromise might come at 23DB of gain with the cap on the FB shunt. I think I have some 330uF caps @ 16V around. I would think they should work fine. About the polarity though, what are your thoughts?
Thanks, Don
Hi Don,
Sounds like you are on the right path. You can experiment with the gain easily enough also.
-Chris
That wasn't lost on me.
Sounds like you are on the right path. You can experiment with the gain easily enough also.
-Chris
I still have the question on polarity for the lytic for the FB shunt. Idea's, suggestions? Would it help to bypass it with a Polyprop of 100nF?
Thanks, Don
Thanks, Don
Hi,Don S said:
if you use NPN or Nchannel input devices then the DC bias at the input is negative with respect to signal ground. If DC offset at the output is zeroed or near zero then the bias at the other side of the input LTP matches the first side.
That is the bias that exists across the cap on the lower leg of the NFB.
If you are unsure, you can measure it. 10mV to 100mV is quite common even when the output DC offset is 0mV.
The bias is modulated by the input signal which can be around 500mVac to 2Vac. The bias pales to insignificance when maximum input voltage is applied.
However a few electrolytic manufacturers specify a 1V or so tolerance to reverse bias. This allows smallish AC signals to be applied without breaking down the insulating film between the plates. But measure the cap bias after installation to be sure you get the best use out of the cap parameters.
Yes, use a film bypass cap. 100nF only becomes effective at high audio frequencies. You might be better trying 1uF or even 2u2F (5mm*7mm*10mm, 5mm pin pitch) if you have room.
I have seen amps deliberately using very high feedback resistor values to allow 9u4F (4u7//4u7F) caps in the feedback line rather than use electrolytics, 110k & 4k3 for a gain of 26.6 (+28.5db) and F-3db of 4Hz.
= a series pair of electrolytics without a biassing voltage at the junction between the plates. That can sound just as bad or worse than any other combination of electrolytics.traderbam said:
Yes, in my opinion, all electrolytics sound very similar in the NFB leg, particularly if they are bypassed with about 1uF of plastic film. But, I adopt near 140mS for this and maybe that hides some of the distortion that the electrolytics may produce.traderbam [/i][B]Why not use a non-polarised electrolytic? [/B][/QUOTE] [QUOTE][i]Originally posted by AndrewT [/i][B] = a series pair of electrolytics without a biassing voltage at the junction between the plates. That can sound just as bad or worse than any other combination of electrolytics. [/B][/QUOTE] [QUOTE][i]Originally posted by traderbam said:
I cannot afford (PCB space and cost) to go all film at this location and I will not go the alternative route of high resistance to make the film caps affordable for just the reason that Roender so forcefully made.
Are you going to question his opinion as well?Ouch
What about the noise from series feedback resistor?
I am not convinced two bipolars are equivalent to a non-polarised cap. From what I've heard, a good non-polar is...quite good. Don't forget, electrolytics have series inductance.
Hi,
yes, ESR and ESL and all the other bits I/we are not competent to measure.
But, bipolar can be made by literally winding two electrolytics together or by double oxidising both surfaces so that operation with either polarity works without failure. Both methods inherently result in a series pair of electrolytics.
The only electrolytics that may avoid part of this problem are a true series pair with the junction polarised to a voltage above the peak signal voltage likely to occur across the capacitor.
That's why I said " or worse".
yes, ESR and ESL and all the other bits I/we are not competent to measure.
But, bipolar can be made by literally winding two electrolytics together or by double oxidising both surfaces so that operation with either polarity works without failure. Both methods inherently result in a series pair of electrolytics.
The only electrolytics that may avoid part of this problem are a true series pair with the junction polarised to a voltage above the peak signal voltage likely to occur across the capacitor.
That's why I said " or worse".
- Status
- Not open for further replies.