Oh and well done by the wayExcellent job.
Thanks
Yes it would be. My issue would only present itself at low drive levels. When driven harder it would allow a decent contact to form that would then persist when the level was reduced. If I turned the amplifier off, so the relays opened, when turned back on again the problem would reappear, only after turning it up did the problem 'go away'.
Of course the problem didn't go away completely, as measured the third order shot up to just below 0.01%, this was when driven at maximum output, so the non-linear part of the contacts was still very much present even though the full drive level was being delivered to the load. If you perhaps remember I had a lot of trouble with a raised third order in the original build of the blameless amps. The culprit was found to be the resistor in the feedback network seeing too large of a voltage swing, it seems that non-linear resistances can cause third order distortion products.
It could also be frequency dependent as I first noticed this when making measurements, I had the tweeter and mid wired in parallel, the tweeter seemed fine, but the mid/bass did not, the low frequency content seemed missing.
I knew relay contacts could become a potential issue, but I did not expect this to occur after only a year. Maybe the relays had been in storage for a while, I don't know, but considering they were fully sealed I didn't expect there to be a problem.
I've had the technique of making good quality PCBs sorted for a few years now, it certainly comes in handy with things like this. As the PCB for the relays had been designed a while ago, all I needed to do was remove the relays and replace them with the fets + opto. The rest was up to Farnell's next day delivery service and of course having an amplifier that doesn't work correctly can be quite an incentive to do something quickly!
There's nothing wrong with perf board mind you, I've used my fair share of the stuff in the past!
Now the question is how long will it take for the other relay based boards to fail? I've got a 4 channel version in another amplifier in another room that was built at the same time, that and a two channel version elsewhere also. If and when I build DC offset/thump protection again I don't think I'll bother with a relay.
The case that the amplifier (and above the DSP/Preamp ) is in has glowing feet. These are RGB LEDs that cycle through a variety of colours. When the protection triggers, or the amplifier is first turned on, the feet turn off, lighting only when there isn't a fault present.
Attachments
Very interesting is this... at the time I never suspected or even thought of the relay as being a potential problem... but your work into this seems to confirm it. It was the midrange driver on mine that seemed to suffer. And like you, it was frequency dependent. My relays are probably around 5 yr old now and I mostly listen at what many would say are low levels most of the time.
Many thanks
Many thanks
That was the thing, I never suspected them either, but something was wrong. I pulled the lid off the amp, got out the scope and dummy load and started looking around. After a bit of probing, the only thing it could have been was the relay.
Do you mean that if no input muting is used the amplifier could be amplifying music when the relays are opened/closed and that this could cause a tiny arc that would hasten the death of the contacts?
Yes, exactly that's why I went for it. Not only that but it uses less power when in operation, doesn't make any noise, has lower overall on resistance then relay contacts do, is much faster in operation and takes up less PCB space.
Four FETs and 1 two-channel opto cost £3.50 per channel, which is far less then a relay spec'd to switch high currents at DC. Of course most of us decide to buy a cheap relay and will make do of replacing the thing if its contacts get scorched when the amplifier breaks. Of course that is assuming the relay manages to do it's job in the first place. I don't relish the thought of having my expensive drive units melted should the worst happen - after all the entire system is active and without a cap in sight, so any of the drivers would be at risk should a fault occur.
Save the relays for the NTC thermistor softstarts. The twin FET relay is the only way for a massive PS DIY amp. I often thought about "how can the OEM's get away with it ?? " . They use current limited trafo's and small main capacitor banks ... nowhere near the massive overkill that DIY'ers end up with (800-1kVA toriods + 50-100Kuf). OUR amps can make a much more massive arc just by design. This is Not RELAY territory.Think industrial...
OS
The twin FET relay is the only way for a massive PS DIY amp. I often thought about "how can the OEM's get away with it ?? " . They use current limited trafo's and small main capacitor banks ... nowhere near the massive overkill that DIY'ers end up with (800-1kVA toriods + 50-100Kuf). OUR amps can make a much more massive arc just by design. This is Not RELAY territory.Think industrial...OS
Well here we go a 6 channel solid state relay based DC protection and turn on thump remover.
This is based around the FR3607 MOSFET produced by International Rectifier and the ASSR-V622 MOSFET driver produced by Avago. The delay and detector is based on the design by Rod Elliott. I'd had Rod's design implemented around some Omron sealed relays, but after a year things started to act up.
A few days ago I decided to measure the distortion performance of my blameless 6 channel, just to check that things hadn't gone out of balance, needless to say one channel had gross 3rd order distortion. At the time I didn't bother to investigate why, then a day or two later I noticed that one of the drivers in the active loudspeakers wasn't producing much sound. Turned out it was the contacts in the relays and a day or two later another one stopped working properly too. I had spotted this thread a week or two ago so it couldn't have come at a better time!
The solid state based relay works great and, here's a distortion sweep of the channel that was previously misbehaving.
As you can see, problem solved
Wow - that was quick and looks neat as well - great job!
Andrew, wrt to your queations/comments about the mosfet devices.
The trench devies can switch in about 100nS easily. however, to do this they need decent gate drive. Looking at the data sheet, I think with the driver circuit (Avago, toshiba, IR), the switching can be well under 100uS. So, I dont think you will have any problems.
For a +-75V rail amp, I would use 200V devices - there are plenty of suppliers on the market.
The trench devies can switch in about 100nS easily. however, to do this they need decent gate drive. Looking at the data sheet, I think with the driver circuit (Avago, toshiba, IR), the switching can be well under 100uS. So, I dont think you will have any problems.
For a +-75V rail amp, I would use 200V devices - there are plenty of suppliers on the market.
Hi Bonsai,
thanks for that speed/delay confirmation.
To all.
I am thinking of these trench FETs for power rail switching (on the amp PCB) in event of fault condition. This allows the fuses to be located back at the source of power, i.e. just after the main smoothing bank.
I wonder if delayed FET switch to ON will eliminate the output glitches that can damage drivers and/or create unpleasant noise? This would allow the front end to start up before the output stage. Any thoughts?
thanks for that speed/delay confirmation.
To all.
I am thinking of these trench FETs for power rail switching (on the amp PCB) in event of fault condition. This allows the fuses to be located back at the source of power, i.e. just after the main smoothing bank.
I wonder if delayed FET switch to ON will eliminate the output glitches that can damage drivers and/or create unpleasant noise? This would allow the front end to start up before the output stage. Any thoughts?
Last edited:
Andrew,
if you switch the rails, you could wait until the voltages across the caps have stabilized, and then switch the power rails to the amp. You can get away with lower voltage mosfets doing this - e.g. 100V on a 75V rails. However, what you should not do is linearly ramp the mosfets because then they will almost certainly fall outside of their SOA. with trench mosfets, you need to switch hard and fast.
if you switch the rails, you could wait until the voltages across the caps have stabilized, and then switch the power rails to the amp. You can get away with lower voltage mosfets doing this - e.g. 100V on a 75V rails. However, what you should not do is linearly ramp the mosfets because then they will almost certainly fall outside of their SOA. with trench mosfets, you need to switch hard and fast.
Thanks
By the way, what do the failure modes of MOSFETs look like? With BJTs the internals tend to short together. If a FET used in this application failed in that way it would of course fail to switch off, which could be potentially lethal.
I don't intend to ramp up. It's quite clear from all the previous posts that fast switching is required and that in turn requires highish currents.
3nF in 100ns seems like quite a bit of current from the gate driver.
Good discussions from everyone. Nice thread to read.
There's a bunch of hardcore power engineers at my workplace - they've looked at high power (> 100A) hot swap applications. They parallel multiple FETs to handle the DC load current but the design goal is such that the startup current (and consequently the SOA as determined by the soft-start period) must be able to be safely handled by just ONE FET.
They have both mathematical and empirical proof (company proprietary). So no matter how many FETs you have in parallel, until they are fully enhanced, assume no current sharing.
There's a bunch of hardcore power engineers at my workplace - they've looked at high power (> 100A) hot swap applications. They parallel multiple FETs to handle the DC load current but the design goal is such that the startup current (and consequently the SOA as determined by the soft-start period) must be able to be safely handled by just ONE FET.
They have both mathematical and empirical proof (company proprietary). So no matter how many FETs you have in parallel, until they are fully enhanced, assume no current sharing.