Is this useful?
An externally hosted image should be here but it was not working when we last tested it.
Yes, an SOA issue most likely. Interesting concept. BTW I was apps support for the 1875, 675, 391, LM12 back in the day. Can hardly believe they are still being used and discussed.
Hello, I hate to revive an old thread but has anyone else built this in a while? I have built several gainclone amps after reading the great article/review on the 47 labs gaincard many years ago. The amp was paired up with a set of Lowther pm2a based horn speakers and the combination was seen as a huge plus by the reviewer, mainly because of the high efficiency of the Lowthers. I experimented with inverting topology and biasing into class a via an 8k resistor to -V, also for buffer I used TVC volume control. I can concur that that amp did pair well with high efficiecy speakers. Real magic there!
My question is in this circuit how critical is the value of the 10u caps on the input. and are these needed if there output caps or 0 DC coming from sources? thanks.
My question is in this circuit how critical is the value of the 10u caps on the input. and are these needed if there output caps or 0 DC coming from sources? thanks.
If everything is ideal, the input capacitors are not needed. The inputs of the chipamps carry a deliberate offset, but of opposite sign. So they cancel out at the input. But some unbalance somewhere will mess up the output offset, so it is better to keep the caps in, especially when an input transformer is used, as was shown in a working example. Transformers are not good in handling DC.
They can be smaller, though. With 10u the LF corner frequency will be around 1 Hz, so you can take a value of 1u for a value of 10 Hz.
Steven
PS This might be my first reply in 5 years or so; I have no time to follow diyaudio actively anymore. I just got a mail from diyaudio, that there was an activity on this old thread and I thought why not.
Maybe also time to change that photo from 2003 in my avatar.
They can be smaller, though. With 10u the LF corner frequency will be around 1 Hz, so you can take a value of 1u for a value of 10 Hz.
Steven
PS This might be my first reply in 5 years or so; I have no time to follow diyaudio actively anymore. I just got a mail from diyaudio, that there was an activity on this old thread and I thought why not.
Maybe also time to change that photo from 2003 in my avatar.
Very high efficiency speakers require very low noise amplifiers. The chipamps are not very low noise. In my view they are not even low noise. Typical Ein of around 20nV/rtHz to 60nV/rtHz for the 3886, compared to 6nV/rtHz to 12nV/rtHz for a good discrete front end.
That's a total variation from best to worst of 20dB, or the difference from 84dB speakers to 104dB speakers. i,e a 104dB speaker on an amp with Ein = 6nV/rtHz will sound like it has the same noise as an 84dB speaker fed from a Ein = 60nV/rtHz
Last edited:
Hi Andrew,
So to get great sound from high effeciency speakers we shouldn't try something like this with a chip amp? I know what I heard with my own ears back in the day, my pm2as on open baffles, paired with a fast sub. My source at the time was better than average digital player which I modded heavily. Mostly redbook was what I listened to. And I powered the amp with batteries. So the noise floor I could hear was very quiet. I didn't have tools back then to do any measurements. And the review that got me started down that path concurs with the approach. The reviewer, who had lived with an Ongaku for years even compared what he heard from the 47labs/Lowther combo to it favorably. The inverted TVC buffered and class a biased version had me listening to all of my music again as the first time. My first gainclone sounded ok thru the same setup but nothing like this newer version. I'm curious to listen to a more appropriately class-a biased gainclone and this seemed interesting, though I no longer have the pm2as
Question is still there thou for this circuit, 10u caps needed on input if no dc from source?
Also Andrew what would a good class-a discreet project be? What I like about chipamps is I can use batteries or maybe SMPS as power.
So to get great sound from high effeciency speakers we shouldn't try something like this with a chip amp? I know what I heard with my own ears back in the day, my pm2as on open baffles, paired with a fast sub. My source at the time was better than average digital player which I modded heavily. Mostly redbook was what I listened to. And I powered the amp with batteries. So the noise floor I could hear was very quiet. I didn't have tools back then to do any measurements. And the review that got me started down that path concurs with the approach. The reviewer, who had lived with an Ongaku for years even compared what he heard from the 47labs/Lowther combo to it favorably. The inverted TVC buffered and class a biased version had me listening to all of my music again as the first time. My first gainclone sounded ok thru the same setup but nothing like this newer version. I'm curious to listen to a more appropriately class-a biased gainclone and this seemed interesting, though I no longer have the pm2as
Question is still there thou for this circuit, 10u caps needed on input if no dc from source?
Also Andrew what would a good class-a discreet project be? What I like about chipamps is I can use batteries or maybe SMPS as power.
If you take a standard power amplifier with an Ein of 12nV/rtHz and give it a gain of 30times, then the output noise is:
Voutnoise = 12nV * sqrt(20000) * 30 = 51uV (-94.9dBW)
Apply that to a 86dB/W speaker and you get an SPL @ 1m that is ~9dB below audibility.
Apply that to a 106dB/W speaker and you get an SPL @ 1m that is ~11dB above audibility.
But the 106dB/W speaker does not need an amp with 30times gain.
Reduce the gain (and recheck the stability margins) to 4 times.
The Voutnoise = 12nV * sqrt(20000) * 4 = 6.8uV (-112.4dBW)
Apply that to your 106dB/W speaker and you get an SPL @ 1m that is 6dB below audibility.
The low gain Power Amplifier is 17.5dB quieter (at the output), than the too high gain version of the same amplifier. (as a check compare 30times to 4times: 30x = +29.5dB & 4x = +12dB, a difference of ~17.5dB).
Just find a good amplifier that has a gain more appropriate to the sensitvity of your speakers. It does not need to have better equivalent input noise. The lower gain gives you an adequate s/n ratio.
If you read the Forum article on system gain. It gives a story on what your system should do and it will give you the same answer as here. Do not put in excessive gain. That just multiplies the noise.
Voutnoise = 12nV * sqrt(20000) * 30 = 51uV (-94.9dBW)
Apply that to a 86dB/W speaker and you get an SPL @ 1m that is ~9dB below audibility.
Apply that to a 106dB/W speaker and you get an SPL @ 1m that is ~11dB above audibility.
But the 106dB/W speaker does not need an amp with 30times gain.
Reduce the gain (and recheck the stability margins) to 4 times.
The Voutnoise = 12nV * sqrt(20000) * 4 = 6.8uV (-112.4dBW)
Apply that to your 106dB/W speaker and you get an SPL @ 1m that is 6dB below audibility.
The low gain Power Amplifier is 17.5dB quieter (at the output), than the too high gain version of the same amplifier. (as a check compare 30times to 4times: 30x = +29.5dB & 4x = +12dB, a difference of ~17.5dB).
Just find a good amplifier that has a gain more appropriate to the sensitvity of your speakers. It does not need to have better equivalent input noise. The lower gain gives you an adequate s/n ratio.
If you read the Forum article on system gain. It gives a story on what your system should do and it will give you the same answer as here. Do not put in excessive gain. That just multiplies the noise.
power chip amps however are not unity gain stable
the LM3886, for instance, is compensated for minimum gain of 10, and the datasheet circuits are all Av 20 or higher
to safely use these at lower signal gain requires "noise gain compensation" - designing with the DS the low GBW min you may have to accept some high audio frequency noise rise from the noise gain compensation circuit
the LM3886, for instance, is compensated for minimum gain of 10, and the datasheet circuits are all Av 20 or higher
to safely use these at lower signal gain requires "noise gain compensation" - designing with the DS the low GBW min you may have to accept some high audio frequency noise rise from the noise gain compensation circuit
Yes, thanks Andrew and jcx. I will look at the datasheets of both the LM3886 and 3875 before deciding on which one to use, gain etc. They are each suited to different ohm specs, example I used the 3875 for 8ohm, I think the reviewer which I spoke of was using 16ohm with the 47labs gaincard(3875?) and the 3886 does good with 8ohm better yet with 4ohm and wouldn't be suitable for 16ohm. Then their is power(VDC) and gain to consider.
All solid state power amplifiers perform better into higher load impedances.
I can't think of any exception.
The national chipamps are current crippled and do not suit 4ohms loading. They require the PSU to be down graded to prevent the chipamps shutting down, or damaging themselves.
I recommend 8ohms, or 12ohms, or even higher, for the National chipamps.
I can't think of any exception.
The national chipamps are current crippled and do not suit 4ohms loading. They require the PSU to be down graded to prevent the chipamps shutting down, or damaging themselves.
I recommend 8ohms, or 12ohms, or even higher, for the National chipamps.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.