Hi Daanve,
it's true that Tom has a commercial interest and eventually may have emphasized some aspects which could help his 'cause' but I state again that I've found his review absolutely fair and his critiques, emphasized or not, are nevertheless based on facts.
It is certainly also true that this type of collaboration has its limits, we are all aware of this but it was nevertheless useful and interesting.
An example of Tom's good faith, IMHO, is at post Modulus 86 or Fremen Edition
A certain amount of inconscious or not bias is human and we take it in account but IMHO Tom has been fair and collaborative.
So I want to thank Tom again for his work and his (appreciated) suggestions. As well, together with George, for the interesting and instructive, if at times lively 😉 , exchange between them.
As you already said, it's on a whole different scale and the small margin goes on covering risks (last GB I could have to make, and pay, boards again for a supplier problem and for the COVID I've spent no less than 30 hours for posting them) and new parts continous evaluation... but I can understand your point. 😉
Tom, don't bother, you know I appreciate and value your work and opinions. 🙂
Absolutely and thanks again for everything. 🙂
Last edited:
It's clear that you've put a lot of effort in. I know from first-hand experience how much work is involved with getting from a one-off hobby project to a fully documented product. And, yep. You do take a risk, even with a group buy.
I've always valued our conversations. Thank you.
No hurt feelings on my end, but at the same time I feel I've gotten my technical curiosity satisfied and really should get back to focusing on my business. I need to figure out where I want to go for the next little while. Thank you for providing a reason for me to procrastinate on writing my business plan. 🙂
Cheers.
Tom
Thanks 🙂
Me too, my pleasure.
Absolutely and good luck for your next projects. 🙂
My test rig is a 2012 Mac Mini, REW software, Focusrite Clarett 4Pre, and non inductive power resistor. My Fe and real Evo measure closer to Tom’s plots. Baseline around 130 dB, spikes up to -110 dB. Power supply and THD effects are level dependent.
Normal output levels are below 200 mW. 80 - 90 dB playback on speakers that are a real 101.5 dB efficient. Plus crossed to a sub at 60 Hz.
What I get is not much more than the loopback measurement. I do not get -150 dB base on loopback, no amplifier, to start with. But THD and noise at -110 dB should be inaudible.
Something is amiss somewhere though. Both the my Evo and FE are silent with ear in the Klipsch La Scala II speakers. That is with the Clarett connected and nothing playing. Tom hears a buzz from his sample.
I need a better test rig to better see the noise from the amps.
But I always look at the 1 watt 1Khz plots JA posts in Stereophile reviews. Most are much higher noise and THD than I get from EVO and FE. He uses a TOTL AP analyzer like Amir.
I subscribe to Stereophile just to look at JA tests of amps, speakers, and dacs. Have one plot that I trust for each.
Normal output levels are below 200 mW. 80 - 90 dB playback on speakers that are a real 101.5 dB efficient. Plus crossed to a sub at 60 Hz.
What I get is not much more than the loopback measurement. I do not get -150 dB base on loopback, no amplifier, to start with. But THD and noise at -110 dB should be inaudible.
Something is amiss somewhere though. Both the my Evo and FE are silent with ear in the Klipsch La Scala II speakers. That is with the Clarett connected and nothing playing. Tom hears a buzz from his sample.
I need a better test rig to better see the noise from the amps.
But I always look at the 1 watt 1Khz plots JA posts in Stereophile reviews. Most are much higher noise and THD than I get from EVO and FE. He uses a TOTL AP analyzer like Amir.
I subscribe to Stereophile just to look at JA tests of amps, speakers, and dacs. Have one plot that I trust for each.
No. I heard a hiss. White noise. It was audible within 30 cm of the speaker, but not at the listening position. That's pretty common in power amps.
Maybe the tweeters in my speakers are more efficient than what you used. Or maybe my ears are more discerning. This is why I rely on objective measurements instead of the subjective experience in my characterization.
If you want to compare objectively, you need to measure with an AC voltmeter. You'll need a meter that can resolve 1 uV RMS or a low-noise amplifier, such as Groner's,
and a decent DMM. If you measure significantly different from 69 uV RMS with the bandwidth limited by a 20 kHz AES17 filter, we can start talking about differences in the circuits. Until then all we have is personal experiences in different environments using different equipment.
Tom
Maybe the tweeters in my speakers are more efficient than what you used. Or maybe my ears are more discerning. This is why I rely on objective measurements instead of the subjective experience in my characterization.
If you want to compare objectively, you need to measure with an AC voltmeter. You'll need a meter that can resolve 1 uV RMS or a low-noise amplifier, such as Groner's,
and a decent DMM. If you measure significantly different from 69 uV RMS with the bandwidth limited by a 20 kHz AES17 filter, we can start talking about differences in the circuits. Until then all we have is personal experiences in different environments using different equipment.
Tom
Last edited:
if you remember, the relay contact in series with the output troubled me since the beginning...
A looong time ago (back in the days when the original MyRef was first presented) I raised the question, suggesting that a different protection scheme which stays completely out of the way when inactive could (and should) be devised.
My idea was (and still is) to have the relay contact operating "in parallel" rather than in series with the output. That would protect the load by short-circuiting the output instead of disconnecting it, while having no effect whatsoever when open. One or more independently operated relays (or some fuses) would have to be added on the power supply rails (in series, before the last by-pass caps to avoid re-introducing nasty side-effects) to protect the amp itself, too.
Of course, the protection circuit and logic would have to be completely re-designed.
DC detection would have to happen right at the 3886 output (before the series resistors) and the protection circuit must either change its "sensitivity" depending on whether the output relay is open or closed or sense (differential) voltage across that resistors instead of the "absolute" voltage to ground (that is, look for DC current rather than voltage).
Needless to say, this will make the whole protection system somewhat more complicated than the existing one. But, IMHO, it would be worth doing.
Granted, just replacing the existing relay with a better one would be much easier... yet it would be only a "band-aid" solution, possibly mitigating but not solving the problem for good.
Last edited:
That type of protection circuit is known as a crowbar. SCRs and TRIACs are commonly used for this because they conduct current continuously once triggered. The main challenge with a crowbar protection circuit is that the crowbar device needs to be able to handle significant dissipated power - at least until the mains fuse blows.
You don't actually need to change the trigger logic much, but you obviously want to make sure you don't get a false triggering of the crowbar during normal music reproduction (or sine wave testing with a 20 Hz signal).
Another approach is to still use a series switch, but to make the switch from devices that don't contribute anything measurable or audible to the signal. There are quite a few circuits floating around on DIY Audio already. I'm sure one of them can be adapted, and possibly even simplified as the MyRef FE has the protection relay in the ground leg of the speaker.
Tom
You don't actually need to change the trigger logic much, but you obviously want to make sure you don't get a false triggering of the crowbar during normal music reproduction (or sine wave testing with a 20 Hz signal).
Another approach is to still use a series switch, but to make the switch from devices that don't contribute anything measurable or audible to the signal. There are quite a few circuits floating around on DIY Audio already. I'm sure one of them can be adapted, and possibly even simplified as the MyRef FE has the protection relay in the ground leg of the speaker.
Tom
Turn on and off transients
The relay may seem like a crude protection solution. But it is there for a very real reason. The LM3886 generates some nasty transients when turned on if configured as a current pump.
No one would live with the pops from your speakers if there was no relay. Your speakers might not live long either.
A crowbar would help, but not eliminate this. And the LM3886 might not last if output was shorted instead of open circuit as it settles down.
This happens very fast. A couple seconds and all is normal.
if you remember, the relay contact in series with the output troubled me since the beginning...
A looong time ago (back in the days when the original MyRef was first presented) I raised the question, suggesting that a different protection scheme which stays completely out of the way when inactive could (and should) be devised.
My idea was (and still is) to have the relay contact operating "in parallel" rather than in series with the output. That would protect the load by short-circuiting the output instead of disconnecting it, while having no effect whatsoever when open. One or more independently operated relays (or some fuses) would have to be added on the power supply rails (in series, before the last by-pass caps to avoid re-introducing nasty side-effects) to protect the amp itself, too.
Of course, the protection circuit and logic would have to be completely re-designed.
DC detection would have to happen right at the 3886 output (before the series resistors) and the protection circuit must either change its "sensitivity" depending on whether the output relay is open or closed or sense (differential) voltage across that resistors instead of the "absolute" voltage to ground (that is, look for DC current rather than voltage).
Needless to say, this will make the whole protection system somewhat more complicated than the existing one. But, IMHO, it would be worth doing.
Granted, just replacing the existing relay with a better one would be much easier... yet it would be only a "band-aid" solution, possibly mitigating but not solving the problem for good.
The relay may seem like a crude protection solution. But it is there for a very real reason. The LM3886 generates some nasty transients when turned on if configured as a current pump.
No one would live with the pops from your speakers if there was no relay. Your speakers might not live long either.
A crowbar would help, but not eliminate this. And the LM3886 might not last if output was shorted instead of open circuit as it settles down.
This happens very fast. A couple seconds and all is normal.
A crowbar would guarantee burnt fuses on the smallest of glitches.. the relay simply escludes any kind of blips. And re-activates after, without harm.
And no, I will never be able to sit down and place dozens of series FETs, inserted AFTER local bypass caps, to escape burnt fuses because of the crowbar..
I would leave that to the brave experimenter, that kind of series of new problems and much extra complexity.
The idea of a MosFet ~simply in place of the relay..I like that, thank You, Tom! I will elaborate and try it.
And no, I will never be able to sit down and place dozens of series FETs, inserted AFTER local bypass caps, to escape burnt fuses because of the crowbar..
I would leave that to the brave experimenter, that kind of series of new problems and much extra complexity.
The idea of a MosFet ~simply in place of the relay..I like that, thank You, Tom! I will elaborate and try it.
The point of the crowbar is that it blows the fuse. But, yes. You do need to make sure that it doesn't turn on during normal operation and survives until the fuse blows when you do need it.
I'm with you, though. I'd rather use a MOSFET in place of the relay.
Tom
I'm with you, though. I'd rather use a MOSFET in place of the relay.
Tom
Properly sized slow-blown fuses before the by-pass shouldn't. But indeed it would be better to use relays.
why dozens? 😕
There are only two power lines. You'd need just two relays (with two contacts each) instead of one.
That is, one relay for the crowbar (which will take care of the turn-on / turn-off transients, too) and another one (controlled independently) with its contacts in series with the two power lines.
When the relays are not powered, contacts on the power lines should be open, while the crowbar should be closed.
At turn-on only the relay on the power lines is activated (closed); after a given delay (covering the turn-on transients), if there's no DC on any of the chip outputs, the crowbar relay is activated (c. open) and the amp is ready to operate.
If something goes wrong and DC is detected, both relays are deactivated simultaneously (=crowbar closed, power open) and will stay so until the whole amp is power-cycled (no automatic rearm). A "fault" LED may (should) be added to signal the error condition.
To avoid driving the amp while the crowbar is closed (e.g., if the amp is powered on while the source is already "playing", or there is a fault upstream which brings DC to the inputs), another pair of contacts of the crowbar relay (or, better, another relay operated in parallel with that one) may short-circuit also the overall signal input while the crowbar is closed(*).
That's it. What's the problem?
(*) a switch forcing the input+crowbar relay(s) to activate may act as a convenient "mute" function, too...
Last edited:
Both Doug Self and Bob Cordell have sections/chapters in their respective books about amp protection. They're worth reading.
Tom
Tom
I think here we are explicitly thinking only about "speaker protection" - there is zero need to protect the amp itself, it is robustly resolved by the designers of the LM3886..
This is why I do not see reason for overcomplications..
Ciao, George
This is why I do not see reason for overcomplications..
Ciao, George
Yep. We are indeed talking about speaker protection. And I agree: No reason for over-engineering it. But I do see a need to engineer a proper solution.
Tom
Tom
what if the DC on the output is caused by a failure (short-circuit) of a 3886 itself?
Though very unlikely, it's not impossible. With a crowbar style speaker protection such a failure can cause severe damage to the amplifier, including the risk of fire.
You need to include some protection against prolonged over-current on the supply lines, too. OTOH, being that a rather unlikely event, also simple fuses could be used for that purpose (that's why I was also suggesting that trivial option).
Perhaps also the usual fuse on the AC line before the power transformer may be enough, though I would feel safer with something (fuse or other device) acting "closer" to the possible short-circuit (that is, on the DC power rails).
That's what fuses are for. 🙂 The crowbar should be designed to ensure that the mains fuse blows if the crowbar ever activates. If a fuse blow can't be guaranteed, the crowbar is not a good protection device for the exact reason you mention.
It would be even more elegant if the protection circuit also turned the power supply off and required power cycling the mains voltage before the supply would turn on again.
Tom
Last edited:
Here (My_Ref) the relay is used as a time delay for output offset isolation. Every time the amp is powered.
This takes a second or two. A crowbar would not work for this function.
This takes a second or two. A crowbar would not work for this function.
True.
On the flip side, a relay may not protect your speaker when it needs to. You can see some spectacular examples of that here: Speaker DC protection with relays
Tom
On the flip side, a relay may not protect your speaker when it needs to. You can see some spectacular examples of that here: Speaker DC protection with relays
Tom
My only example with a My_Ref was at least 20 years ago. A connection in the feedback loop opened up while music was playing. The output voltage went to almost negative full rail.
The poor single driver speaker was slapped with over 30 volts dc. Sounded like a shotgun shot in the house. I was upstairs and ran down. Wife expected to see parts of my diy gear burning.
At the least I expected to find the Lowther PM7A had a smoked voice coil. Found the bad solder joint and touched with iron.
Powered up and relay closed. Speakers went back to playing music. For years.
Became a believer in the Shrack relays and the output protection circuitry. Even then those were 600 - 700 USD voice coils.
The poor single driver speaker was slapped with over 30 volts dc. Sounded like a shotgun shot in the house. I was upstairs and ran down. Wife expected to see parts of my diy gear burning.
At the least I expected to find the Lowther PM7A had a smoked voice coil. Found the bad solder joint and touched with iron.
Powered up and relay closed. Speakers went back to playing music. For years.
Became a believer in the Shrack relays and the output protection circuitry. Even then those were 600 - 700 USD voice coils.