| Vikash |
Phew, what an effort - but worth it! I'm on top of the world at the moment. My first chasis (hell, first metal working project), and first fully working two channel GC is complete!
This one uses Brian's basic PCB with a 300VA toroid (2x 20v) and Sfernice P11 10k linear pot.
After running it for a couple of hours, I took it to a friends this evening and hooked it up to his Martin Logan Ascent's and Shanling CDT-100 to see what it could do. It didn't appear to have any problems driving the Logans at moderate levels. There was clearly more bass, perhaps a touch more clarity noticeable on the highs, and better overall imaging than the Denon AVC-A1SR that we used as reference.
Still many tweaks to go, but so far I'm a very happy bunny. :D
(Must thank PD et al. in the metal working thread for all the help and advice, and of course Brian. :up: )
A bit of a David and Goliath thing going on here...
 |
|
|
| Peter Daniel |
I have to point out that proper placing of the amp (feet, supporting platform, etc.) is quite influencial on the sound. Putting it on top of Denon is not optimal solution, I would think.
I did some further experiments with a chassis, and it seems like tightening the bolt on transformer is important too. I preferred to tight it only slightly, only as much as needed for firm support.
You might also notice that removing top cover from the amp, has more influence than removing plastic jackets from electrolytics. The sound is smoother and seemingly more air can be perceived. I still keep the top on, but don't tight the screws at all.
Nicely done, BTW. I like the centrally positioned knob and warmth of side panels.;) |
|
|
| JasonL |
| quote: | Originally posted by Peter Daniel
I have to point out that proper placing of the amp (feet, supporting platform, etc.) is quite influencial on the sound. Putting it on top of Denon is not optimal solution, I would think.
|
that is not a amp that would be a cd player / dvd player : O) |
|
|
| Vikash |
The feet are 40mm square blocks of acrylic roughly 15mm thick. I plan to put the LED's inside the feet, I just hope they don't melt. I don't believe my ears will tell me any different if I remove/loosen the cover, adjust the toroid bolt, or remove the cap jackets - but I'll give it a go nonetheless ;) I'm looking forward to a thorough listening comparissons once the premium kit version has been finished.
A shot of the internals:
 |
|
|
| Taco |
Very nice chasis.
PD: are those chips so sensitive for microphonics, that feet/ support platform/ transformer mounting has influence on the sound? |
|
|
| Nuuk |
Yes, nice work Vikash. That's a pro job allround! ;)
Just one thought though! :att'n:
You measured all your resistors to a decimal place to ensure that you balanced the channels and then you used a pot instead of a stepped attenuator! :( :xeye: :eek: :bawling: |
|
|
| Upupa Epops |
| To Nuuk : How big is error ( difference of gain between channels ) by using 1 % resistors ? ;) |
|
|
| Vikash |
| quote: | | You measured all your resistors to a decimal place to ensure that you balanced the channels and then you used a pot instead of a stepped attenuator! |
Very true :clown: but I felt I was taking on enough for a fist amp without adding diy attenuators to the list. I took some quick measurements of the channel difference on the pot and and it was not good, so I will almost certainly need to change this down the line. At least the pot will be easier to change than the resistors though ;)
In fact I found an odd quirk whereby you could only hear output in one speaker until you reach a certain volume. This was still at a low volume but I haven't traced the cause yet.
On a last thought, I don't know if there's enough space in the chasis for an attenuator :rolleyes: |
|
|
| MikeB |
| quote: | Originally posted by Peter Daniel
You might also notice that removing top cover from the amp, has more influence than removing plastic jackets from electrolytics. The sound is smoother and seemingly more air can be perceived. I still keep the top on, but don't tight the screws at all.
|
Hmm, if your amp is that sensitive to such details, maybe you should
consider that your amp is slightly oscillating ?
I never experienced these details having impact on the sound with
stable circuits. Only the unstable ones showed such symptoms...
Or you have incredible detailed ears compared to mine ? :scratch:
Mike |
|
|
| Nuuk |
| quote: | | In fact I found an odd quirk whereby you could only hear output in one speaker until you reach a certain volume. This was still at a low volume but I haven't traced the cause yet. |
Is that linear pot law-faked, and of so, with what value resistor?
I believe that law-faked pots can be a bit tempermental at the 'ends'. ;) |
|
|
| MikeB |
| quote: | Originally posted by Vikash
In fact I found an odd quirk whereby you could only hear output in one speaker until you reach a certain volume. This was still at a low volume but I haven't traced the cause yet.
|
My experience from solidstate amps is, that this can be caused by
an incorrect inputfilter. But i am not familiar with the BrianGT...
BTW, very nice chassis ! I hope i get mine nearly that nice...
Mike |
|
|
| Vikash |
| Pot is not law faked yet. Could it be that the channel difference is so far out that is causing it I wonder? I only recall hearing this on the Logan's however. |
|
|
| Nuuk |
| quote: | | Pot is not law faked yet. |
It's got to be worth a try. Solder a 1K5 between the out and ground pins. :smash: |
|
|
| Vikash |
| It's on the todo list, but isn't the taper of a volume control just a convenience feature to match human volume matching - i.e. there's no quality to be gained from it right? |
|
|
| d_anders |
| Looks very nice, but what do you use for heat sink? |
|
|
| Vikash |
| The chips are connected to a plate of copper 4mm thick which is inturn connected to the the aluminium base which is again 4mm thick. I have not tested thoroughly yet, but I found that the chips do not get past warm to the touch, and this is therefore adequate, or even superfluous ;) |
|
|
| Nuuk |
| quote: | | It's on the todo list, but isn't the taper of a volume control just a convenience feature to match human volume matching - i.e. there's no quality to be gained from it right? |
Yes, it is to enable a linear pot to mimic the action of the log pot. ;) |
|
|
| falcott |
| quote: | Originally posted by Vikash
I don't know if there's enough space in the chasis for an attenuator :rolleyes: |
Huh?? Seems to me like you have a HUGE amount of room, especially if you shift the amps slightly. Much more space than what I am attempting at the moment - trying to squeeze 2 GC amps with those big Black Gates on perf boards, a buffer valve and attenuator into a 120mm cube (separate psu, though).
Nice tidy job, by the way! :) |
|
|
| Nuuk |
I agree with Falcott - there should be room for a stepped attenuator in there! ;)
I have just shoe-horned one into one of my 4 inch diameter pods together with a buffer. |
|
|
| Vikash |
Regarding the channel balance problem at low volume, the pot is the culprit. I took some more measurements and the channels seem balanced to within 3% except for at very low levels where it is hugely innacurate. Simply a faulty pot, but otherwise 3% is better than I was expecting (it's rated 10%).
| code: |
left right diff. %
0.40 0.40 0.00 0.0
94.30 11.40 -82.90 87.9
411.00 328.40 -82.60 20.1
800.50 711.50 -89.00 11.1
2002.00 1946.00 -56.00 2.8
2552.00 2523.00 -29.00 1.1
3066.00 3086.00 20.00 0.6
4543.00 4591.00 48.00 1.0
5510.00 5597.00 87.00 1.6
6137.00 6208.00 71.00 1.1
6980.00 7125.00 145.00 2.0
7846.00 8040.00 194.00 2.4
8940.00 9180.00 240.00 2.6
9380.00 9610.00 230.00 2.4
|
|
|
|
| Vikash |
I finished the premium components amplifier today but left the pot out. DC on one of the channels is 120mV without a resistor on the input, and down to about 70mV with ~14k on the input.
What makes the DC offset on one channel different to the other when all components are the same between channels?
The DC measurement is the same regardless of whether or not there is a load on the output. Am I measuring it correctly? |
|
|
| Peter Daniel |
| The DC offset depends on a chip itself. The latitude here varies on average +/- 80mV and in few instances I observed as much as 100mV of offset (2 chips per 100). This with 22k on the input. |
|
|
| Vikash |
I just looked at the circuit diagram again and noticed there's already 22k between input and ground. duh.
Ok so I'm getting 120mV with 22k on the input. Change the chip? Or is this an acceptable amount? |
|
|
| Peter Daniel |
I can send you a new chip.
There is no problem with using it as it is now, as with volume decrease the offset will go down too, but in a long run it's better to change for a better performing chip. |
|
|
| Vikash |
| Thanks Peter, that's a kind offer. Fortunately, I have about 8 spare ;) I wish there was a way of measuring the feasbility of a chip without having to desolder the old chip and solder the new one on with the fiddly feedback resistor before seeing if the DC offset is acceptable... |
|
|
| Peter Daniel |
| Actually there is a way to do that. If you have an unassembled Brian's board, solder all the parts in except for the chip. The mounting holes (for a chip) are pretty tight and they can work as kind of socket. Just insert the chip in, make sure the pins are tight against holes and measure it. ;) |
|
|
| carlosfm |
| quote: | Originally posted by Nuuk
It's got to be worth a try. Solder a 1K5 between the out and ground pins. :smash: |
And then measure DC-offset again.:bawling: |
|
|
| UrSv |
| quote: | Originally posted by Vikash
Regarding the channel balance problem at low volume, the pot is the culprit. I took some more measurements and the channels seem balanced to within 3% except for at very low levels where it is hugely innacurate. Simply a faulty pot, but otherwise 3% is better than I was expecting (it's rated 10%).
--snip-- |
Not faulty. Normal. |
|
|
| Vikash |
| quote: | Originally posted by carlosfm
And then measure DC-offset again.:bawling: |
Hi Carlos, was this not mentioned just to law fake the 10k linear pot? I've removed the pot from both amps so now there is only 22k from input to ground.
Also, I've used 16R from channel ground on the PCB to chassis ground. I was unsure whether this was sufficient? |
|
|
| Vikash |
I've just measured the basic amp again this time removing the extra resistors I had on the input and now the DC offset measures 78mV and 100mV.
Carlos, the closest resistor I have at the moment is 2k4r which if I add between the signal and ground pins on the board reduces the offset to 35.26mV and 12.5mV :D |
|
|
| Vikash |
| quote: | Originally posted by UrSv
Not faulty. Normal. |
Hi UrSv, 87% out is normal and not faulty when its rated at 10%? I've recieved a replacement from RS and will see if that is as far out... |
|
|
| Peter Daniel |
| quote: | Originally posted by Vikash
the closest resistor I have at the moment is 2k4r which if I add between the signal and ground pins on the board reduces the offset to 35.26mV and 12.5mV :D |
In minimized GC circuit the offset depends on input resistance (from +IN to ground). In case of 22k feedback resistor, I wouldn't go for more than 22k input shunt value. In some case I was having as much as 300mV of offset with input resistance of 50K.
When you choose lower values of input resistance (less than 22k), the offset value will go lower as well. On some chips it crosses 0 V reading and goes into positive (or negative ) offset voltage figure. |
|
|
| UrSv |
| quote: | Originally posted by Vikash
Hi UrSv, 87% out is normal and not faulty when its rated at 10%? I've recieved a replacement from RS and will see if that is as far out... |
IMHO, Yes. Desirable?. No. The first part where the pot starts using the resistive layer usually with certain pots will be way off right when it starts. As soon as you're just a tiny bit into the surface things get better. |
|
|
| Nuuk |
| quote: | | And then measure DC-offset again. |
Could you please elaborate Carlos? :confused: |
|
|
| Vikash |
| quote: | Originally posted by UrSv
IMHO, Yes. Desirable?. No. The first part where the pot starts using the resistive layer usually with certain pots will be way off right when it starts. As soon as you're just a tiny bit into the surface things get better. |
The replacement pot although better than the previous (error is only half now), is stilll way out as you said :up:
Peter, are you saying that its ok to leave the caddock 22k on the board and add another smaller resistor (perhaps 1k5r) between signal in and signal ground? |
|
|
| carlosfm |
| quote: | Originally posted by Nuuk
Could you please elaborate Carlos? :confused: |
Yes.
In NI, with the "standard" schematic, these chips produce more DC-offset.
It really depends on the input impedance (including the pot).
Don't forget, the pot is in parallel with the 22k resistor (from +IN to Ground).
DC will even change with the pot position. An op-amp as an input buffer would avoid this and you could tune input impedance for lower DC.
And, an unpredictable issue, there are no two equal chips.:bawling:
With higher values for Rf and Ri you could use higher values of input impedance.
To avoid high values of DC-offset, sometimes very different from channel to channel, you need to use the Ci caps (Look at Fig. 1 on the LM3875 datasheet).
Good electrolythics bypassed with quality poly 0.1uf caps will bring DC-offset to a couple of mv, without affecting the sound. |
|
|
| Peter Daniel |
| quote: | Originally posted by Vikash
The replacement pot although better than the previous (error is only half now), is stilll way out as you said :up:
Peter, are you saying that its ok to leave the caddock 22k on the board and add another smaller resistor (perhaps 1k5r) between signal in and signal ground? |
And I've heard some comments before that linear pots have better matching;)
I would actually recommend leaving 22k Caddock (on the board) when adding smaller resistor. It should influence the sonics in more positive way. I noticed that having Caddock in shunt position somehow increases the impression of 3-dimentionality, much better than Vishay, for instance. |
|
|
| UrSv |
| quote: | Originally posted by Peter Daniel
And I've heard some comments before that linear pots have better matching;)
I would actually recommend leaving 22k Caddock (on the board) when adding smaller resistor. It should influence the sonics in more positive way. I noticed that having Caddock in shunt position somehow increases the impression of 3-dimentionality, much better than Vishay, for instance. |
True. It's thanks to the linear pot that he was only 87% off with the first one... |
|
|
| carlosfm |
With cheap pots you have to buy a bunch of them.
Log or linear, the problem is the same.
Measure them all, and select the one(s) that have lower impedance mismatch between the two tracks(channels).
Then... it's not cheap.:bawling: :clown: |
|
|
| TaaJ |
If the cost of a worthwhile switch alone wasn't about 150%-200% of a Potentiometer, I'm sure more people would be making Attenuators.
Personally with my first GC I will probably use a Pot, and then when I get better at understanding and building amps I will give it to a friend and build another one with a buffer, attenuator etc.
By that time I'd hope to have a dedicated pre and DAC anyways.
Someone could make ALOT of money if they started making switches at the quality of say Elma but at half the cost. |
|
|
| Nuuk |
| quote: | | Don't forget, the pot is in parallel with the 22k resistor (from +IN to Ground). |
I was assuming that it was isolated by the input cap. :confused: |
|
|
| carlosfm |
| quote: | Originally posted by Nuuk
I was assuming that it was isolated by the input cap. :confused: |
If you use the input cap, yes.
Not everybody uses it.
The input cap will only protect from DC from the source.
With NI, I would rather use the Ci cap. and have 2~5mv DC-offset.:cool: |
|
|
| Nuuk |
| quote: | | With NI, I would rather use the Ci cap. and have 2~5mv DC-offset. |
But I guess you would rather go with inverted anyway! ;) |
|
|
| carlosfm |
| quote: | Originally posted by Nuuk
But I guess you would rather go with inverted anyway! ;) |
Yes.:D |
|
|
| Cappy |
Regarding the Ci cap that Carlos mentions, has anyone tried it on their Gainclone?
I might be interested in trying this with Brian's PCB. Carlos, should I stick with the 22 uF cap in that case?
Thanks. |
|
|
| carlosfm |
22uf is fine if you use 1k for Ri.
If 680 ohms, go for 33uf.
Bypass with 0.1~0.22uf poly.
Yes, I tried with and without that cap.
With the cap I have 4mv on one channel, 5mv on the other one.:D
Without the cap, I don't remember, it was long ago...
But it was much higher than this.
I used 33uf + 0.15uf poly. |
|
|
| Vikash |
I've just read on the gc glossary that to bypass is to offer a low impedance route for high frequency noise. I just want to make sure this means connecting the 0.1uf in parallel with the 33uf?
I'll give the Ci cap a go too ;) |
|
|
| carlosfm |
| quote: | Originally posted by Vikash
I've just read on the gc glossary that to bypass is to offer a low impedance route for high frequency noise. |
:eek:
Why use low ESR caps, then?
:bawling: |
|
|
|
