noise floor etc.
Hi,
first I have to ask you for accepting my apologies, because some posts before I was very wrong concerning the impedance relations between a tube amp output and the UcD input.
I believe JohnW posts should be taken into account, when thinking about linear regulators for the UcD preamp stages. Did he use BG or other "audiophile" caps to reach these noise levels?
A try to calculate this:
Assume a "bad" bypass cap with 1 (Ohm) effective impedance. The AD8620 has a PSRR of about 80dB at 1kHz. Assuming further, it works from 12V rails into a 1k load, the maximum out current will be 12mA, giving a 12mV ripple on rails. This will result in 12mV*(-80dB)=1.2µV PSRR related distortion. The 6nV/sqrt(Hz) input noise, converted to a 20kHz range is 850nV eff or about 5µVpp, above the 1.2µV PSRR distortion. I would say, this is not significant. Am I right?
Regards, Timo
Hi,
first I have to ask you for accepting my apologies, because some posts before I was very wrong concerning the impedance relations between a tube amp output and the UcD input.
I believe JohnW posts should be taken into account, when thinking about linear regulators for the UcD preamp stages. Did he use BG or other "audiophile" caps to reach these noise levels?
A try to calculate this:
Assume a "bad" bypass cap with 1 (Ohm) effective impedance. The AD8620 has a PSRR of about 80dB at 1kHz. Assuming further, it works from 12V rails into a 1k load, the maximum out current will be 12mA, giving a 12mV ripple on rails. This will result in 12mV*(-80dB)=1.2µV PSRR related distortion. The 6nV/sqrt(Hz) input noise, converted to a 20kHz range is 850nV eff or about 5µVpp, above the 1.2µV PSRR distortion. I would say, this is not significant. Am I right?
Regards, Timo
Hi Robin,
Indeed a good idee to upgrade the UcD modules with a DC-dection circuit.
However it will be in the next batch revision
But in our powersupplies, what we will offer soon, we will put here a DC protection.
Cheers,
Jan-Peter
www.hypex.nl
Indeed a good idee to upgrade the UcD modules with a DC-dection circuit.
However it will be in the next batch revision
But in our powersupplies, what we will offer soon, we will put here a DC protection.
Cheers,
Jan-Peter
www.hypex.nl
Just a thought: If you have a relay at the mains and you have DC at the output, how long will it take before you have reached a safe voltage level, in case of disaster? Let's say that you have a conventional PS with 10000-47000 uF, 70 volts. Isn't this enough to fry your speakers? Maybe it's better to have the relays at the power rails instead? The obviuos disadvantage is that the relays must be able to handle much more current but this may not be a real problem.
It's tipped here that MOSFET's with a photovoltaic driver from IR could be used instead of relay(s) for DC protection:
http://www.geofex.com/Article_Folders/ampprot/dcprot.htm
What do you say about it? It would be easy to driver with some DC level sensor (RC -> diode -> NPN -> NPN -> res -> MOSFET driver) etc., same for negative polarity).
Also regarding the differential inputs, which is the best location for volume control (low-noise potentiometer, step attenuator, DS1802 etc.)? Should it also be dual to compensate for the impedance for the (-) input?
http://www.geofex.com/Article_Folders/ampprot/dcprot.htm
What do you say about it? It would be easy to driver with some DC level sensor (RC -> diode -> NPN -> NPN -> res -> MOSFET driver) etc., same for negative polarity).
Also regarding the differential inputs, which is the best location for volume control (low-noise potentiometer, step attenuator, DS1802 etc.)? Should it also be dual to compensate for the impedance for the (-) input?
As JP noted, the relay goes between the large storage caps and the UcD modules so you only get the local 470uF discharging into your speaker. The relays actually handle lower current. In a class D amp, input current is always less than output current (of course it's multiplied by the number of channels you have on one psu).peranders said:
Protection Relays
You cannot effectively put relays in the DC rails. I've done that already and had to move the relays to the wires between the power transformer and the bridge. If you try to protect the amplifier by switching off the DC, the relays will stick. To disconenct a DC circuit which has failed and is sucking a lot of juice, you need a really big relay and even then it may stick. When you disconnect voltage, when there is a lot of current flowing, the relay will arc over as the contact are pulled apart.The current will keep flowing or the relays may merely just stick together when the relay closes.
With AC flowing through the relay, the arc is quenched 100 or 120 times per second when the voltage falls through zero.
I have made this mistake myself and had to move the relays from the amplifier board to the power supply board. Save yourself an expensive re-design and switch the AC leads to the bridge and not the DC rails.
Also, in the trading post section I have some toroidal power transformers for sale that are suitable for the UcD180.
You cannot effectively put relays in the DC rails. I've done that already and had to move the relays to the wires between the power transformer and the bridge. If you try to protect the amplifier by switching off the DC, the relays will stick. To disconenct a DC circuit which has failed and is sucking a lot of juice, you need a really big relay and even then it may stick. When you disconnect voltage, when there is a lot of current flowing, the relay will arc over as the contact are pulled apart.The current will keep flowing or the relays may merely just stick together when the relay closes.
With AC flowing through the relay, the arc is quenched 100 or 120 times per second when the voltage falls through zero.
I have made this mistake myself and had to move the relays from the amplifier board to the power supply board. Save yourself an expensive re-design and switch the AC leads to the bridge and not the DC rails.
Also, in the trading post section I have some toroidal power transformers for sale that are suitable for the UcD180.
More on relays
Yes, by switching the AC lines feeding the bridge, you are discharging the filter caps into your speaker, if you do nat have a separate relay on the speaker. However, I found this happens anyway because with even as little as 220µFd of capacitance on the load side of the relay, the contacts have welded themselves shut. In one power amp I build, I have both a relay on the speaker and a DPST relay on the AC leads to the bridge. The speaker relay saves the speaker, the AC disconnect relay prevents fire.
Having two relays is not a lot more money and it prevents letting all the magic smoke out of your amplifier if there is a failure.
Yes, by switching the AC lines feeding the bridge, you are discharging the filter caps into your speaker, if you do nat have a separate relay on the speaker. However, I found this happens anyway because with even as little as 220µFd of capacitance on the load side of the relay, the contacts have welded themselves shut. In one power amp I build, I have both a relay on the speaker and a DPST relay on the AC leads to the bridge. The speaker relay saves the speaker, the AC disconnect relay prevents fire.
Having two relays is not a lot more money and it prevents letting all the magic smoke out of your amplifier if there is a failure.
Hi,
you could always use forklift type of relay (contactor) to disconnect DC. They have magnetic arc extinguisher built around contacts.
Seriousley, I was thinking before about Bruno's reccomendation. Use of normally connected relay would prevent welding contacts together at turn on, but would certainly weld them together at turn of if protection circuit has operated prior to turn off. There are some relays with DC rated contacts, but i think it is better to use another solution. Bypassing relays with mosfets is useable but not very practical solution. Using fuses and SCR crowbar is another, again not very elegant one. I think that proposed two relay solution is technically valid, but I really have no idea about the sonic impact of the output relay.
Best regards,
Jaka Racman
you could always use forklift type of relay (contactor) to disconnect DC. They have magnetic arc extinguisher built around contacts.
Seriousley, I was thinking before about Bruno's reccomendation. Use of normally connected relay would prevent welding contacts together at turn on, but would certainly weld them together at turn of if protection circuit has operated prior to turn off. There are some relays with DC rated contacts, but i think it is better to use another solution. Bypassing relays with mosfets is useable but not very practical solution. Using fuses and SCR crowbar is another, again not very elegant one. I think that proposed two relay solution is technically valid, but I really have no idea about the sonic impact of the output relay.
Best regards,
Jaka Racman
I've used a capacitor across the relay contacts to prevent these sort of problems (with an R across the cap to ensure it gets discharged). At the instant the contacts open, the arc is eliminated by transiently diverting the current into charging up the cap. (Like in an old-type vehicle distributor contact breaker).
Sounds like that must be sufficient. We mustn't forget that on both sides of the relay there's an electrolytic. Left, the 10mfd, right the 470uF of the amplifier. So the amount of inductive energy stored in that loop can't be very much.Ouroboros said:
Dan, in the instances you quote where the relay contacts were welded together upon opening, was there a cap on both sides?
What about when we should use a Powerfet like the IRF540 and IRF9540 as switches?
A P-channel in the positive rail and a N-channel in the negative rail, the gates are connected with a resistor to GND. With some transistors we pull the gates to zero voltages in a case of emergency. The gates ofcourse protected with a 12V zener to avoid to high voltage.
Cheers,
Jan-Peter
A P-channel in the positive rail and a N-channel in the negative rail, the gates are connected with a resistor to GND. With some transistors we pull the gates to zero voltages in a case of emergency. The gates ofcourse protected with a 12V zener to avoid to high voltage.
Cheers,
Jan-Peter
Hi,
I do have experience with arcing in relays. When the arc starts, there is no way of cutting it off otherwise than cutting the power off. Plastic parts of the housing melt down so they usually prevent relay from closing the contact again after the power has ben restored. I tend to agree with Bruno that capacitors on both sides would prevent arc. After all this is equivalent to having a cap right across the relay.
But welding of contacts upon turn on is a sure thing. I once used a small relay to disconnect one of the PCB boards in the system. Board consumption was 100mA at 5V and it had a single 1uF ceramic capacitor for blocking. Of course I tested the relay with 30000 on/off cycles. When the production started, all the relays welded after 200-1000 cycles. I measured turn on current and it was 60A pk 1us wide half sine. That was enough to weld the contacts on the relay which was otherwise rated for 1A. I also performed additional testing on the original relay. It had failed after 62000 cycles. So much about single sample testing.
Jan-Peter, you can use fets or bipolars as switches which are then bypassed by relay. If you turn them on at the same time, caps will be charged before the relay even closes. This is what I did to solve problem with sticking relay.
Best regards,
Jaka Racman
You
I do have experience with arcing in relays. When the arc starts, there is no way of cutting it off otherwise than cutting the power off. Plastic parts of the housing melt down so they usually prevent relay from closing the contact again after the power has ben restored. I tend to agree with Bruno that capacitors on both sides would prevent arc. After all this is equivalent to having a cap right across the relay.
But welding of contacts upon turn on is a sure thing. I once used a small relay to disconnect one of the PCB boards in the system. Board consumption was 100mA at 5V and it had a single 1uF ceramic capacitor for blocking. Of course I tested the relay with 30000 on/off cycles. When the production started, all the relays welded after 200-1000 cycles. I measured turn on current and it was 60A pk 1us wide half sine. That was enough to weld the contacts on the relay which was otherwise rated for 1A. I also performed additional testing on the original relay. It had failed after 62000 cycles. So much about single sample testing.
Jan-Peter, you can use fets or bipolars as switches which are then bypassed by relay. If you turn them on at the same time, caps will be charged before the relay even closes. This is what I did to solve problem with sticking relay.
Best regards,
Jaka Racman
You
In my own amps I have the psu connected through the "normally on" contacts. Turning it on when the psu is charged would kill it.
The relay needs to be energised to disconnect the two sides. This happens only in case of fault. So, as long as the amplifier is working normally (ie over the amplifier's entire lifetime), the relay is never operated.
When a DC fault occurs, the coil is energised. The main contact opens and the alternate contacts are used to latch it on (coil is powered from the "left" side through the alternate contact). As long as the plug isn't pulled, the relay stays latched. When the plug is pulled, the relay drops off (comes on actually) only when the psu caps have discharged to a fairly low voltage.
The relay needs to be energised to disconnect the two sides. This happens only in case of fault. So, as long as the amplifier is working normally (ie over the amplifier's entire lifetime), the relay is never operated.
When a DC fault occurs, the coil is energised. The main contact opens and the alternate contacts are used to latch it on (coil is powered from the "left" side through the alternate contact). As long as the plug isn't pulled, the relay stays latched. When the plug is pulled, the relay drops off (comes on actually) only when the psu caps have discharged to a fairly low voltage.
Some (very few) "ordinary" relays have been characterized with safe operation area curves for dc use. Look through enough data sheets and you'll eventually come across one. If I remember right, one model I once used had a 10A rating at 12Vdc (and 240Vac), but the rating started falling at somewhere around 18 volts and was down to 100mA at 48 volts.
I used the relay for reverse voltage protection on a product's 48 volt battery input. The relay wasn't allowed to close if the input voltage was negative. The relay was followed by a mosfet in series with it to actually make and break the load. The normal downstream load was around 50 volts at 6 amps, yet the circuitry only let the relay switch well within its very limited safe operating area.
Even though most relays haven't been rated for beyond 12 or 24 volts dc, by dissecting and comparing a few representative samples, one could select unrated types with similar construction to the ones with the desired dc ratings - and then verify this with some life testing under both normal and fault conditions. Under fault conditions a lifetime of only a hundred cycles or so would probably be enough.
I think the material, thermal mass, and breaking distance of the contacts all play a roll in their suitability for dc use. Of course, there are specialty relays designed to break large dc fault currents while opening into high voltages. These are usually designed to magnetically blow out an arc. On its way from the contacts to the terminals, load current flows through an open coil of a few turns situated such that, under fault, it sets up a strong magnetic field. This field forces the arc out and away from the contacts so that it must pass in a long serpentine path through a large arc chute (many parallel non conducting louvers). The long path quenches the arc.
Back to the problem at hand: Jaka, since the circuit in question already has capacitors in place to snub stresses at contact opening, your suggestion of adding a mosfet in parallel to the relay seems like the simplest way to go to me.
Regards -- analog(spiceman)
Output Relays
Output relays have been in use for 30+ years and are about as benign a part as there is. The alternative is a crowbaw across the output with a big triac to discharge the caps after the relay disconnecting the AC side of the bridge has opened. This would need to be a 25A device at least and may not survive. However, it would only fire if the output transistors were already blown so what's an extra $3.00 part if it saved your speakers.
By the way, when I did try relays on the DC rails, I tried a 470 ohm resistor and a .1 cap in parallel with the contacts and with a new relay. I figured the 470 ohm resistor would pre charge the cap on the ICE module, as the ICE module draws almost no current on the main power rails until the mute is brought high. However, because the DC voltage was still different on the two sides, this only increased the number of turn ons before the relay welded to about 20 from 4.
Switching the DC rails with relays, I delermined can never be reliable unless very large relays are used. Relays that are ale to reliable break the DC are as big as the modern power amp stages.
Now, if you use FETs instead, there are a couple problems. First, you will end up with a protection circuit as complex as your output stage. Second, the low but not zero Rds resistance of the FETs in series with the filter caps and outside the feedback loop of the amplifier will have a subtle but definite effect on the sound of an amplifier. Whether a good or bad effect is unknown. Most changes to the sound quality are generally considered bad. Regardless, the impedance of the FETs will have to be added to the output impedance of the amplifier when calculating damping factor and will likely result in looser bass and a sloppier sound.
As well, these FETs will generate as much heat as a Class D output stage. If the amp is Class AB, this is a relatively small addition. If the amp is CLass-D, it will double the amount of heat that needs to be dissapated.
Finally, these are solid state devices and if there is an output stage failure, the currents that are being tossed around are quite large. It is likely that a failure in the output stage may result in failure of the pritection device as well.
I believe the cost of the extra heatsinks and the cost of the FETs and the possible effect on the sound quality versus using either a second relay or a 3 or 4 pole relay makes the relay option better over all.
You can buy a 3 or 4 pole relay. Use 2 poles for the AC leads to the bridge and one pole or even two in parallel if you have them for the speaker.
Output relays have been in use for 30+ years and are about as benign a part as there is. The alternative is a crowbaw across the output with a big triac to discharge the caps after the relay disconnecting the AC side of the bridge has opened. This would need to be a 25A device at least and may not survive. However, it would only fire if the output transistors were already blown so what's an extra $3.00 part if it saved your speakers.
By the way, when I did try relays on the DC rails, I tried a 470 ohm resistor and a .1 cap in parallel with the contacts and with a new relay. I figured the 470 ohm resistor would pre charge the cap on the ICE module, as the ICE module draws almost no current on the main power rails until the mute is brought high. However, because the DC voltage was still different on the two sides, this only increased the number of turn ons before the relay welded to about 20 from 4.
Switching the DC rails with relays, I delermined can never be reliable unless very large relays are used. Relays that are ale to reliable break the DC are as big as the modern power amp stages.
Now, if you use FETs instead, there are a couple problems. First, you will end up with a protection circuit as complex as your output stage. Second, the low but not zero Rds resistance of the FETs in series with the filter caps and outside the feedback loop of the amplifier will have a subtle but definite effect on the sound of an amplifier. Whether a good or bad effect is unknown. Most changes to the sound quality are generally considered bad. Regardless, the impedance of the FETs will have to be added to the output impedance of the amplifier when calculating damping factor and will likely result in looser bass and a sloppier sound.
As well, these FETs will generate as much heat as a Class D output stage. If the amp is Class AB, this is a relatively small addition. If the amp is CLass-D, it will double the amount of heat that needs to be dissapated.
Finally, these are solid state devices and if there is an output stage failure, the currents that are being tossed around are quite large. It is likely that a failure in the output stage may result in failure of the pritection device as well.
I believe the cost of the extra heatsinks and the cost of the FETs and the possible effect on the sound quality versus using either a second relay or a 3 or 4 pole relay makes the relay option better over all.
You can buy a 3 or 4 pole relay. Use 2 poles for the AC leads to the bridge and one pole or even two in parallel if you have them for the speaker.
Relay
Also, if you use something to switch off the rails, either on the DC or AC side, this circuit has to be latching, with a manual reset switch or by cycling the power off and on. This is beacuse if you do not have a latch and when the circuit trips, the DC offset will go away after the caps are discharged and the protection circuit will come on again and the cycle will repeat. Basically, the protection circuit will oscillate at a low frequency.
Therefore, you need some sort of latch that is powered by a separate power source. Powered from the AC secondary before the disconnect relay and protected either by a small fuse or a self resetting polyswitch.
Also, if you use something to switch off the rails, either on the DC or AC side, this circuit has to be latching, with a manual reset switch or by cycling the power off and on. This is beacuse if you do not have a latch and when the circuit trips, the DC offset will go away after the caps are discharged and the protection circuit will come on again and the cycle will repeat. Basically, the protection circuit will oscillate at a low frequency.
Therefore, you need some sort of latch that is powered by a separate power source. Powered from the AC secondary before the disconnect relay and protected either by a small fuse or a self resetting polyswitch.
I have some experience with switching the low voltage AC side of the powersupply. The disadvantage will be that there will be some energy in the capacitors, when the relais is triggerd by the DC protectioncircuit. Ofcourse we will use a form of latching and resetting by switching the powersupply off.
Personal this would be for me the best solution, I do not like at all to put a relais in the loudspeakeroutput. For shure not with the UcD180
In the past we have tested two different brands relays, Omron 16A and Zettler 16A. The Omron went down after 2000-4000 times switching the Zettler after >20.000 times!!! So we now only use Zettler
Cheers,
Jan-Peter
www.hypex.nl
Personal this would be for me the best solution, I do not like at all to put a relais in the loudspeakeroutput. For shure not with the UcD180
In the past we have tested two different brands relays, Omron 16A and Zettler 16A. The Omron went down after 2000-4000 times switching the Zettler after >20.000 times!!! So we now only use Zettler
Cheers,
Jan-Peter
www.hypex.nl
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.