Trimpots are really only good for a limited number of operations. I believe a standard spec. was a minimum 200 rotations, end to end and I've watched a few guys exceeding that with endless tweaking and fiddling to get whatever they were looking for just so, not just in in one session, but in several attempts, every time a circuit or component change was made. So I've had to replace pots that were no longer correct in value or even working at all. See if you can measure a voltage that changes as you rotate the control arm and check that this a smooth, continuous change. I don't expect any problem but you need to be certain if and why something that should be happening, isn't.
YES, leaky caps could play havoc, especially ageing, small electrolytics. Film caps should be OK for a lifetime. I haven't checked the schematic though, to see what would likely occur, not that I'm an expert with this design as it confused me too, back in 1983.
YES, leaky caps could play havoc, especially ageing, small electrolytics. Film caps should be OK for a lifetime. I haven't checked the schematic though, to see what would likely occur, not that I'm an expert with this design as it confused me too, back in 1983.
Last edited:
The trim pot I used for the test was a brand new 1k 10 turn.
Set to midway before inserting in circuit. Each "end" of the pot is connected to the emitters of Q1 and Q2.
Wiper connected to the 47K where the two 270r also connect.
adjusting the pot whilst measuring the output voltage across the speaker terminals
Input shorted and no speaker connected.
Whilst turning the pot, the voltage appears to reduce a little, but when I stop turning , the voltage crept back up to about where it was before.
This is why I asked the question about some sort of "servo" action that was compensating for the now out of balance diff pair.
When turned fully to the end, ( you can hear the gears clicking as they jump ) the output voltage suddenly drops dramatically to around 18mV.
In this position, the 270r resistor is effectively shorted out, whilst the other 270r effectively has 1k across it in parallel.
Again to the question of possible "servo" action, once this happens the "servo" can no longer compensate, and we actually get a change in the offset voltage.
Set to midway before inserting in circuit. Each "end" of the pot is connected to the emitters of Q1 and Q2.
Wiper connected to the 47K where the two 270r also connect.
adjusting the pot whilst measuring the output voltage across the speaker terminals
Input shorted and no speaker connected.
Whilst turning the pot, the voltage appears to reduce a little, but when I stop turning , the voltage crept back up to about where it was before.
This is why I asked the question about some sort of "servo" action that was compensating for the now out of balance diff pair.
When turned fully to the end, ( you can hear the gears clicking as they jump ) the output voltage suddenly drops dramatically to around 18mV.
In this position, the 270r resistor is effectively shorted out, whilst the other 270r effectively has 1k across it in parallel.
Again to the question of possible "servo" action, once this happens the "servo" can no longer compensate, and we actually get a change in the offset voltage.
Then the trimpot wiper must be dropping off the resistance element at that end of its travel. That shouldn't prevent it from working up to that point though. There likely is something that seems like a slow servo action when you alter the offset though. You attempt to change the steady DC potential at the amplifier output which is controlled by another DC potential derived from the input stage current sink . The amplifier fights the change as it settles to its new setting and this may come as surprise. I don't recall this type of action with my ETI5000 but then, I haven't needed to check it for many years.
Looking at the ETI477 schematic alex mm just posted, the resistance and trimpot values seem to all be different. I realise that it is probably SuzyJ's version and the semis are different types too but I would expect it to work similarly to ETI477 ( i.e. the amplifier boards in the ETI5000 project)
Looking at the ETI477 schematic alex mm just posted, the resistance and trimpot values seem to all be different. I realise that it is probably SuzyJ's version and the semis are different types too but I would expect it to work similarly to ETI477 ( i.e. the amplifier boards in the ETI5000 project)
I second the good advice already offered by Janusz and Ian. Yes with MOSTFETs the more bias the better and ideally at least 100mA per MOSFET pair really sweetens up the sound by lowering crossover distortion. But the it will run VERY hot. If the amp is going back into use in a studio where it will be on for long periods or all the time I'd keep the bias at the original setting. If it's going to be used for serious audiophile listening sessions in a home environment then set the bias as high as you dare and turn it off when you finish listening 
Try it and see if you think the sonic improvement is worth it.
Regarding your offset problem - it is in only one channel and the other channel is ok yes? In that case you have a good reference to compare with. If you compare all voltages carefully between the two boards you should be able to find the culprit.
I'd definitely be replacing the input devices with BC546, all the ceramic caps with good quality NPO types, and changing C2, C3 and C14 to metal film polypropylene (5mm pitch) or polystyrene (both available from Rockby in Mebourne). The input and feedback electrolytics definitely should be replaced. It is possible that one or more MOSFETs are suspect - I had one fail in one of the amps I repaired. Luckily I had an original Hitachi device spare. Check the voltage across every source resistor - they should all be similar. Look for signs or excessive heat around any of the output devices. Checking for oscillation with a scope would be advisable if you can.
Regarding the mod to equalise dissipation in the VAS stage, a BF or equivalent device wired in series between Q7c and Q5c with its base grounded would be an elegant solution technically. But I have not tried it or looked at how you'd do it on the PCB. And note you're moving the power dissipated from the diff amps (which is the goal) into the new common base device so it will need heatsinking also so would need to mount on the existing VAS heatsinks somehow. I wouldn't try it at this stage - concentrate on getting it going properly first and leave this for tweaking later if you feel like trying it - it might not make any difference to the sound and is not related to your offset issue.
Try it and see if you think the sonic improvement is worth it.Regarding your offset problem - it is in only one channel and the other channel is ok yes? In that case you have a good reference to compare with. If you compare all voltages carefully between the two boards you should be able to find the culprit.
I'd definitely be replacing the input devices with BC546, all the ceramic caps with good quality NPO types, and changing C2, C3 and C14 to metal film polypropylene (5mm pitch) or polystyrene (both available from Rockby in Mebourne). The input and feedback electrolytics definitely should be replaced. It is possible that one or more MOSFETs are suspect - I had one fail in one of the amps I repaired. Luckily I had an original Hitachi device spare. Check the voltage across every source resistor - they should all be similar. Look for signs or excessive heat around any of the output devices. Checking for oscillation with a scope would be advisable if you can.
Regarding the mod to equalise dissipation in the VAS stage, a BF or equivalent device wired in series between Q7c and Q5c with its base grounded would be an elegant solution technically. But I have not tried it or looked at how you'd do it on the PCB. And note you're moving the power dissipated from the diff amps (which is the goal) into the new common base device so it will need heatsinking also so would need to mount on the existing VAS heatsinks somehow. I wouldn't try it at this stage - concentrate on getting it going properly first and leave this for tweaking later if you feel like trying it - it might not make any difference to the sound and is not related to your offset issue.
Hello alex mm - nice work on your PCB design. Have you built and tested it? How is the sound?
I note the schematic does not quite match the PCB - I see more output devices for starters. And there are some component changes from the original design. The diodes in series with the supply rails to the early stages are unusual - I recall these being used in the Electronics Australia Pro Series 1/3 designs but nowhere else that I've seen. Since that was a poor sounding design I would be suspicious of them. Have you tried shorting them out and comparing the sound? If you really think the filters are needed (and in my view since it's a fully differential design and should have inherently high PSRR they probably aren't), 10R + 1000uF would be preferable. You lose less volts this way too, which is already an issue with lateral MOSFETs.
I note the schematic does not quite match the PCB - I see more output devices for starters. And there are some component changes from the original design. The diodes in series with the supply rails to the early stages are unusual - I recall these being used in the Electronics Australia Pro Series 1/3 designs but nowhere else that I've seen. Since that was a poor sounding design I would be suspicious of them. Have you tried shorting them out and comparing the sound? If you really think the filters are needed (and in my view since it's a fully differential design and should have inherently high PSRR they probably aren't), 10R + 1000uF would be preferable. You lose less volts this way too, which is already an issue with lateral MOSFETs.
I think Alex's electronic design was drawn up by Suzyj when she designed and built her own boards in SMT. Her design was modular and mounted on the heatsink. Unfortunately, this also made it desirable for commercial manufacture, so, it was subsequently pirated by a US company and modules were produced and sold there for some years without royalty or even attribution.
alex mm contributed his fine and professional through-hole board too, based, I believe, on suzy's schematic as this would have suited more DIY constructors.
It might be difficult now to identify the significant changes without following through the thread to completion but her pro. lab test results were exemplary and drew a lot of attention from the experts here. ETI 5000 MOSFET Power amp
alex mm contributed his fine and professional through-hole board too, based, I believe, on suzy's schematic as this would have suited more DIY constructors.
It might be difficult now to identify the significant changes without following through the thread to completion but her pro. lab test results were exemplary and drew a lot of attention from the experts here. ETI 5000 MOSFET Power amp
Yes i mentioned this. That is the point where the end of the resistance track and the wiper become ZERO ohms.
This is why it shorts out the 270ohm Q1 emitter resistor.
This is the only time that the output offset voltage makes any REAL change.
Surely there is something quite out of balance if, on the input diff pair, when one emitter resistor is zero ohms and the other is 270ohms and the output voltage is getting closer to zero V but just not quite there yet ???
And yes, I even tried swapping Q1 and Q2 around. = no difference.
I understand that there will be a bit of "lag" between adjustments and the resulting change on the output. I was patient and waited minutes between tweaks.
Quite a different circuit to the original ETI477 circuit.
This issue is present on both boards. pretty much identical.
Input and feedback caps replaced. (they test perfect) = no difference.
Q1 &Q2 replaced (matched hfe, matched Vbe ) = tiny difference. ( i have tried about 10 pairs so far. difference in offset voltage has dropped from about 80mV (original 550's) down to about 50mV (with any of the replacement 550 pairs)
Measuring the voltage across the source resistors (0.22ohm) all voltages are about the same. From memory they were about 0.02 volts across each with a variance of about 0.003 between resistors. This is not unusual as these things would be lucky to have 10% tolerance in their resistance spec.
Measuring voltages across other components as per the original ETI477 (not some hybrid frankenamp) schematic, they are all pretty much exactly as defined by David Tilbrook.
Measured with Fluke 77 after a 30 minute warm up. with inputs shorted and outputs open.
Power supply voltages measure ~ +48.2 volts and -48.5 V
And just to be clear. The actual circuit that I am working off, which is the original Tilbrook ETI477 schematic is attached as a PDF to this post.
Just for some background.
I purchased this amp ( and a series 5000 pre-amp) from a local guy who constructed them back in 1986 from (Altronics) kits.
When I got them home I obviously lifted the lids and was astonished to find damn near pristine condition. Nice big heatsinks on the VAS BF's.
Perfect front heatsink.
Perfect case.
Really neatly laid out wiring.
Very heat PCB soldering.
All components oriented identically, across both boards.
Mosfets only have the slightest patina.
Actually very close to what I would consider "commercial" build quality.
All in all, a VERY nice original example ( same with the pre-amp )
So much better than what I built back in '85.
This is why I am interested to get this fine example to be as good as it can be.
This is a piece of Aussie DIY electronics HiFi history from a very highly regarded engineer and is very well worth preserving in its original spec.
I'm not looking to modify it into a 0.000000000001% THD, ultra low noise, theoretically perfect straight wire with gain amp.
Now, I understand 50mV of DC appearing on the output (with no input) is possibly an acceptable level for this "type" of amp. ( ie, non specialist DIY gear ) (this is no $10,000 Holton ! )
The original offset voltage spec was stated to be -/+25mV (page 32, ETI January 1981) but maybe too low to measure !!
I just feel that with a bit of tweaking it could be better, say 20mV. Obviously this goal is a lot more involved than that.
Maybe I should just button it up, wrap it in plastic and put it on the shelf in the shed and put it up for sale in another 30 years.
If it's a good and desirable asset and having it original, standard, unmodified and untweaked is best, then leave it that way. Gilding the lily by fine tuning it to your own preferences won't often add value, unless your buyer is lazy, or really expects a late model, high end commercial product.
At one time, there was a lot of (silly) comment on the 'net about bass speaker drivers being damaged or the response being adversely affected by the voice coil being in the order 0.5mm from their ideal rest location, central in the magnetic gap. 'Too bad that typically, there is also a mechanical offset anyway, even when there is no DC offset voltage.
Obviously, passive crossovers use capacitor coupling for the tweeter and midrange drivers, so there is nothing to have a DC issue with there, but I have even read criticism of DC offsets in passive systems at HF and that can only have come from the wrong end of the author
At one time, there was a lot of (silly) comment on the 'net about bass speaker drivers being damaged or the response being adversely affected by the voice coil being in the order 0.5mm from their ideal rest location, central in the magnetic gap. 'Too bad that typically, there is also a mechanical offset anyway, even when there is no DC offset voltage.
Obviously, passive crossovers use capacitor coupling for the tweeter and midrange drivers, so there is nothing to have a DC issue with there, but I have even read criticism of DC offsets in passive systems at HF and that can only have come from the wrong end of the author
Looking at that 477 schematic, I see a "shield" connection that could be used instead of a direct signal earth connection and was probably intended to isolate input signal grounds from power ground currents and reduce hum - perhaps in a utility or PA amplifier. I don't recall that feature in ETI5000 because the 477 was also a separate project and I wasn't about to build it with just a DIY PCB and shopping list.
Notice that the virtual ground is established at the midpoint of electrolytic caps connected to the power rails and I suspect DT wasn't happy with that adaptation nor other mods made by ETI staff.
About 10 years ago, I had a similar discussion here and via PM, with member sparkey, who had an ailing hifi system and blamed his ETI 5000. He wanted to fix his bias setting problem too and I suggested he buy a 200R trimpot, and wire the trimpot ends directly to Q1,2 emitters with the wiper to R11 - ditching R10,13. Sparkey had a few problems but he said he was surprised and happy with the amplifier then and celebrated with a new pair of tower speakers, so I now suspect there was a design or component problem there. I really don't like cheap knock-off, multi-turntrimpots either. They are really annoying and rubbery in their adjustment, so where sensible, I now revert to single-turn cermet types. Expensive? Yes, the real deal can be well over $2 and you could probably buy a box full of Chinese multi-turns for the same price but I wouldn't rely on the stability over say, 10 years, which is the very least you could expect from a permanent setting adjustment.
Notice that the virtual ground is established at the midpoint of electrolytic caps connected to the power rails and I suspect DT wasn't happy with that adaptation nor other mods made by ETI staff.
About 10 years ago, I had a similar discussion here and via PM, with member sparkey, who had an ailing hifi system and blamed his ETI 5000. He wanted to fix his bias setting problem too and I suggested he buy a 200R trimpot, and wire the trimpot ends directly to Q1,2 emitters with the wiper to R11 - ditching R10,13. Sparkey had a few problems but he said he was surprised and happy with the amplifier then and celebrated with a new pair of tower speakers, so I now suspect there was a design or component problem there. I really don't like cheap knock-off, multi-turntrimpots either. They are really annoying and rubbery in their adjustment, so where sensible, I now revert to single-turn cermet types. Expensive? Yes, the real deal can be well over $2 and you could probably buy a box full of Chinese multi-turns for the same price but I wouldn't rely on the stability over say, 10 years, which is the very least you could expect from a permanent setting adjustment.
This shield connection was definitely part of the "series 5000" project and described in the text;
"The power amplifier will regard as a valid input any voltage difference between the input and the input earth terminals. So any hum present on this
earth will be treated as an input and amplified accordingly. In order for the
hum level to be inaudible from a 100 W power amplifier it must be at least
90 dB below the full output voltage, which is around 0.9 mV. Since the
voltage gain of the ETI-477 is approximately 23, the equivalent input signal
voltage is 0.9 mV/23 _ 39 uV! It is clear that even a minute hum level at
the input will produce an audible hum at the output. To overcome this problem
the input earth is isolated from the 0 V track on the circuit board by the 10 ohm resistor R3, shown on the ETI-477 circuit diagram .
I've attached an extract from the ETI article
Interested to know where this virtual earth point should be connected to ??
It's based on the attached version with the exception of the transistor instead of the 18-20V zener. Purpose was to use 3 pairs of Exicons (as I have a number of these) and make the pcb more versatile. This version has not yet been tested.
cheers,
Thanks for the ETI5000 article details. A 10R "ground lift" resistor is now quite common in all powers of grounded audio amplifier systems. Nowadays, the grounding of domestic appliances, including audio systems, is treated as a serious safety matter but you can find how to implement a proper earth loop breaker that offers a high degree of protection, at ESP, figure 5 in this article
Last edited:
My PhD supervisor built a pair back in the eighties and claimed you could tune out 2nd order distortion by moving the ground connection point.
Regarding value, it’s an eighties kit amp, built with the cheapest components Jaycar or Altronics could source at the time, on single-sided boards. That they still work at all is testament to the design skills of DT. I doubt they’ll improve in value, especially given that those of us who lusted after these things in the early ‘80’s are now in our late forties and fifties with plenty of disposable cash now. They’re at their peak.
I think the heatsink casting for the ETI5000 was pretty cool though.
OP: I have a stash of BF469s and BF470s, and I’m in Perth, if you want some. I can also measure your poor amps to death.
Thanks for the offer Suzy, is a PM ok ?
I have had a bit of a more closer read of the original article and have a couple more things to look at. And I will try replacing the BF's with matching pairs of equivalents.
One specific question though, if anyone can give a definitive answer.
The gate resistors (100ohm).... what is the actual effect of changing them to 220ohm ???
I have had a bit of a more closer read of the original article and have a couple more things to look at. And I will try replacing the BF's with matching pairs of equivalents.
One specific question though, if anyone can give a definitive answer.
The gate resistors (100ohm).... what is the actual effect of changing them to 220ohm ???
Leinster Lad - if your offset is similar and high on both boards and stays after replacing the obvious suspect components I wonder if it is actually a power supply/grounding issue.
The amp you have may have been built neatly, but is it wired according to the article? I would check all the wiring very carefully and make sure it matches the part 3 article. Two of the amps I repaired were not wired properly but were otherwise neatly constructed. There could be some issues with the BF devices in the VAS, but it seems odd that both boards would have the same problem. One other thing - where did you buy the BC550s from? If from Jaycar/Altronics they could be counterfeit and might already be breaking down due to running so close to their maximum Vce.
Having said that, if the offset is only 80mV this is not going to bother any speakers nor cause any other issues so might not be worth worrying about.
I second Suzy's comments re cheap and nasty components in the local kits. Also the output devices are obsolete, so the thing is living on borrowed time...
The amp you have may have been built neatly, but is it wired according to the article? I would check all the wiring very carefully and make sure it matches the part 3 article. Two of the amps I repaired were not wired properly but were otherwise neatly constructed. There could be some issues with the BF devices in the VAS, but it seems odd that both boards would have the same problem. One other thing - where did you buy the BC550s from? If from Jaycar/Altronics they could be counterfeit and might already be breaking down due to running so close to their maximum Vce.
Having said that, if the offset is only 80mV this is not going to bother any speakers nor cause any other issues so might not be worth worrying about.
I second Suzy's comments re cheap and nasty components in the local kits. Also the output devices are obsolete, so the thing is living on borrowed time...
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.