carlosfm said:
I would be interested to see how you create the delay, I gather that it is not difficult, but you never know until you get the elbow grease out (Aussie slang for getting down to the nitty gritty).
LM3875 also comes out as LM3876 with mute - as I far as can see, any delay solution on 3886 should work exactly the same on 3876.
Joe R.
PS: You guys know anybody who has tried 3886 and not felt it inferior to 3875?
Joe Rasmussen said:
Check the LM3780 datasheet, at page 15, "Mute mode".
This datasheet is more complete than most.
Joe Rasmussen said:
Yes, I know somebody.
Me.😀
I'm thinking about trying this with my LM1875 amp,but thought I'd ask a question before I get the parts. The data sheet for this chip says that it's "stable for gains of 10 or greater". I wondered if there is an "optimal" gain setting for this chip, or if I would be better to stick with the suggested gain of 10?
mrskinny said:
It's interesting to note that the original Gaincard has more gain than NS's recommended gain of 10 (20dB), with a setting just above 30dB. 'Stereophile' actually published the gain in a review. I would say, aim at a gain of at least 30 (very close to 30dB) and if the results are anything like LM3875, use around 45 (33dB). More gain, yes, but also lower feedback and feedback virtually zero above 200KHz where the 90 degree phase lag goes off (bad, really) and you want closed loop gain equal open loop gain. Actually, this is similar to classic feedback theory that an amplifier needs to be equal to or less than unity gain before feedback becomes positive due to the phase lag.
But there is one huge proviso to making this work, and the Gaincard also uses this solution. The two 1000uF caps is right on the power chip pins and the 0V point at which they are joined must become the point at which the speaker return lead is also joined. As the chip is essentially open loop above 200KHz, the HF PSR is poor and since these chips use the neg rail as a signal reference, spurious HF currents are possible. Use fast low ESR 1000uF/50V electros (Panasonic FC is a good choice) and wiring suggested (basically copied from the Gaincard and its sibling), and no problems.
This has been repeated, but useful to reiterate. Repetition is the Mother of Retention.
Joe R.
carlosfm said:
That's a very high gain for who (like me) wants to make a power amp.
I have what seems a negative suggestion: Use an input attenuator, voltage divider (isn't that what a volume control is anyway and we use them all the time). Use resistors types to suit your own taste. Seems crazy to slug gain, whether before or after you have created it, but what can I say, if it works... and noise shouldn't be a problem as that has already been demonstrated.
Let your ears become the final arbiter. If it works... it works!
Joe R.
Joe Rasmussen said:
The National Tech notes for the LM4780 give a simple slow-start circuit for the 4780 and I don't see why it wouldn't apply to the 3886. I have one on my PCB but have yet to fiddle with it to test the mute timing.
See page 15
National 4780 PDF
Mmmm...
It really doesn't appeal to me to attenuate and amplify again.
That's what the pre is doing too.
But then again, thinking... if I use a good pot (like an Alpha 50k log😉 ) on the input, this amp could do as a power amp or an integrated.
Kevin Haskins said:
Yes, it should work with any National chip with the Mute function.
Kevin Haskins said:
Kevin, I will have to do that too.
I'm thinking in 30~40 seconds turn-on delay.
Who makes it first posts it.
Deal?
carlosfm said:
Kevin, I will have to do that too.
I'm thinking in 30~40 seconds turn-on delay.
Who makes it first posts it.
Deal?
I'd be happy to but my PCB is limited to a digital file at the moment.
I could use your input though!
Kevin's LM4780 PCB
carlosfm said:
Hi Carlos
Since RM goes to neg rail and the Mute Off needs 0.3mA min and likely 0.5mA to be on the safe side, then RM max value is 68K (using nearest standard R value) @ -35V rail . So varying delay can only be done by varying cap value, right? The current gradually increases through R as cap is charged, this also gradually lessens the mute, unlike a relay that just comes On. This is the kind of thing that cannot be done with a slide rule, but only hands-on. At least that is my working method with situs like this, where things are not On/Off like a timing relay.
So you cannot say you are going to get x seconds of delay as the mute is graduated in time. At 0.1mA you get -20dB, that may be muted not but mute, if you know what I mean. I would say that 0.01mA and -70dB is pretty close to it. But there is a transition time between 0.01ma, 0.1mA and 0.3mA. So you have to make the whole time delay longer than it initially looks.
Possibly a fet current source that is gradually biased in some way, that eventually achieves target 0.5mA and plateaus there, with smaller timing cap on the gate side? That way achieve more than one minute before the plateau is reached, but effective muting is less than that.
I need to mull over that one, but it should be possible? I've got a 35V power suppy, might jig something up and get it working safely first.
There is also something else I've been working on: If using cathode follower with split +/- rails, then from the cathode to ground, put a 10-15K resistor in series with 100uF/50V Bipolar electro. This dampens the turn-on and stops the neg rail to be hit (this is the buzz they are hearing - the ripple of the neg rail) and clips the amp at the same time. With RC damping, this is slowed down, the cone still wobles a bit but clipping (and buzz) avoided - largely silent now. About 10 seconds later, it hits the positive rail and a short burst of buzz off the pos rail. Again the RC damping avoids the clip/buzz. Gradually the cap charges and effectively only the R load is seen as parallel with load after the coupling cap. The follower bias itself normally once bipolar cap is no longer sucking DC current.
But going over to 3876 or 3886 opens up other possiblities. As you can see, this whole topic is of interest to me.
Joe R.
PS: Which soocer team do you follow. Benfica, Sporting, Porto... ?
OK, you Euro guys are probably just waking up, but I have been busy.
This shows a delay technique based on a current source. The IRF830 is a Hexfet that doesn't start conducting current until about 2.5V applied to gate relative to source. This is used to good effect here. The 10uF is charged by 13M6 (2x6M8) until 1M5 resistor causes the voltage at the source to plateau. The 470R is adjusted to give near 1mA at the plateau. This value is arrived at by disconnecting 10uF (no delay) and 470R value is adjusted to give 1mA across it. Then reconnect 10uF and it should be operable.
Some test showed that it took 18 secs approx before 2.5V was reached across 1M5 and hence gate bias. During this period the fet is in cut-off mode. As bias voltage rises slowly above 2.5V, the current, seen as voltage across 470R (or whatever value you end up using). In my case it took 39 secs to reach 0.3mA - and at this point there will be no attenuation in the amp chip. It took 31 secs to reach 0.1mA and at this value the attenuation in the 3876/3886 is still -20dB, in fact around -23dB according to NS's graph.
I put the above together using a 36V power supply, but it should work quite well. The 10uF is critical, it must be very low loss. If longer delay is required, then the value could be increased to 22uF, but again, the cap needs to be really good.
Just as test, I have some 39uF/63V Panasonic FC. This is a premium brand and I was curious how it would stack up.
Where the previous periods were 18 secs, 31 secs and 39 secs with 10uF,with 39uF they were 69 secs, 2 minutes 10 secs and 2 minutes 53 secs.
Almost 3 minutes before full non-attenuation is achieved, that is just a bit too long. So premium Panasonic FC 22uF should fit in about the middle. The leakage was as good as non-existent - this was check by comparing bias voltage across 1M5 with and and without cap. The FC cap made no difference. Hey, these are really good!
Joe R.
An externally hosted image should be here but it was not working when we last tested it.
This shows a delay technique based on a current source. The IRF830 is a Hexfet that doesn't start conducting current until about 2.5V applied to gate relative to source. This is used to good effect here. The 10uF is charged by 13M6 (2x6M8) until 1M5 resistor causes the voltage at the source to plateau. The 470R is adjusted to give near 1mA at the plateau. This value is arrived at by disconnecting 10uF (no delay) and 470R value is adjusted to give 1mA across it. Then reconnect 10uF and it should be operable.
Some test showed that it took 18 secs approx before 2.5V was reached across 1M5 and hence gate bias. During this period the fet is in cut-off mode. As bias voltage rises slowly above 2.5V, the current, seen as voltage across 470R (or whatever value you end up using). In my case it took 39 secs to reach 0.3mA - and at this point there will be no attenuation in the amp chip. It took 31 secs to reach 0.1mA and at this value the attenuation in the 3876/3886 is still -20dB, in fact around -23dB according to NS's graph.
I put the above together using a 36V power supply, but it should work quite well. The 10uF is critical, it must be very low loss. If longer delay is required, then the value could be increased to 22uF, but again, the cap needs to be really good.
Just as test, I have some 39uF/63V Panasonic FC. This is a premium brand and I was curious how it would stack up.
Where the previous periods were 18 secs, 31 secs and 39 secs with 10uF,with 39uF they were 69 secs, 2 minutes 10 secs and 2 minutes 53 secs.
Almost 3 minutes before full non-attenuation is achieved, that is just a bit too long. So premium Panasonic FC 22uF should fit in about the middle. The leakage was as good as non-existent - this was check by comparing bias voltage across 1M5 with and and without cap. The FC cap made no difference. Hey, these are really good!
Joe R.
Good morning, no, good evening, Joe!
I use just this simple circuit (with LM3875 to connect the input to ground):
An externally hosted image should be here but it was not working when we last tested it.
Franz
Joe Rasmussen said:
Joe, (I say, asking as a neophyte), in your DIY tube GC schematic, there is already a 10k (from pins 3 and 8) to the -ve rail. Am I to understand that I can add the cap in series there, so that it quietens the warm-up noise?
Cheers!
Franz G said:
Hi Franz,
looks like the T-network thread is becoming unravelled for the moment.
But I am interested in making a turn-on delay for my VBIGC. Can you describe your circuit, because I have to admit that it has gone over my head...
😕
Thanks!
If I understand correctly, we are talking about two approaches to avoiding power up noise/damage to speaker drivers.
One invloves muting (Joe's circuit above) or isolating the speakers with relays as in the various speaker protection modules.
The other is grounding the input of the amp as in the circuit Franz has submitted.
Correct me if I am wrong but I want to undesratand all this and expect other less experienced builders will too! 😉
One invloves muting (Joe's circuit above) or isolating the speakers with relays as in the various speaker protection modules.
The other is grounding the input of the amp as in the circuit Franz has submitted.
Correct me if I am wrong but I want to undesratand all this and expect other less experienced builders will too! 😉
Joe Rasmussen said:
Yes, that's what I think that will happen, from reading the datasheet.
Maby it will work fine, or maby not as we expect.
It's a thing to try.
Thanks for spending your time with this.
Joe Rasmussen said:
Benfica
- Status
- Not open for further replies.