Ci and DC offset ??

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.

decky

Member
2006-09-12 6:10 am
Hi guys,

I have just completed my first gainclone 3886 kit (Brian GT). It is a dual mono configuration with two trafos 160VA each (25V supply).

I decided to ommit the Ci capacitor to avoid any capacitance in the signal path. The measured DC offset without load is 97mV on one chanell and 120mV on the other. The sound is pretty good but I keep wandering is this DC offest good enough and is the 23mV difference between chanells something I should worry about. Would the DC offset and the noise floor be any lower with 47uF Ci?

Second question - how dangerous is driving 6ohm speakers with this amp considering that the supply is chosen for 8ohms. Is it only thermal dissipation I should worry or there are some other issues.

Thanks and sorry if I am raising same old stupid questions but I could not find answers by seraching.
 

decky

Member
2006-09-12 6:10 am
I thought that a rule of thumb (for the worst case scenario) is 10mV * Gain (33 in this case). This is a non-inverting kit and Brian suggested DC offset <100mV - I am more worried about the 25mV differential.

Any suggestions about the 6ohm speaker load?
 
Those values are for no-signal state. However, you should expect much more DC leaking in with signal present, so I think omitting decoupling capacitors in this design is not a good idea.
Instead of lossy and distorting 47 uF electrolytics, you can use 4,7 uF MKP (1/10th of design value for elecotrlytics) without losing the bottom end response.
As for the impedance matching, it depends on how your speakers behave and what is their minimal impedance. Most of 8 Ohm loudspeakers will go down near to 4-5 Ohm on some frequencies and this is taken into account in amplifier designs. It is reasonable to suspect that a 6 Ohm loudspeaker would go even lower and this can significantly drain your output stage and possibly kill it.
 

decky

Member
2006-09-12 6:10 am
I really have to be completely anal now. Lets go back a step. DC offset defines the noise floor if I understand everything properly. The Ci is a filter capacitor to serve as a high pass filter together with R3 (680 ohm in Brian's layout) adjusted to very low frequency - 47uF and 680ohm leads to f=1/(2PI*R3*Ci)=5Hz - therefore 4.7Hz will filter below 0.5Hz so might be not as efficient. The LM3886 data sheet aims at 7.2Hz with Ri and Ci combination of 1kohm and 22uF respectively.

Also I am not sure that I can physically fit an MKP capacitor on the board. Would tantalum or at least NP electrolytic be any better?

Also, bare in mind that the sound from the amp is pretty awesome. I have a bit of noise only on my CD modified output stage. Everything else works as a charm.

Regarding the speaker - here is the impedance curve

http://www.stereophile.com/standloudspeakers/241/index4.html

It does not look too bad :confused:
 

decky

Member
2006-09-12 6:10 am
Nuuk said:
Well short the input and make sure you have a load (10 ohm resistor will do) on the outputs and then measure again. ;)

Ok Ok - right channel 17.3mV - left 16.8mV - but which method is more relevant.

Also considering that I have bit of a noise when using CD player is it really worth while cramping a big MKP cap on such a small board?
 

decky

Member
2006-09-12 6:10 am
It is a funny one - initially I had a bit of a hum (60Hz) that I solved with a ground breaker circuit (diodes, cap and a resistor in parallel). That worked well. But with the CD/SACD it is a bit different - it is a Sony SCD-XB940 modified by JLTi (Allen Wright). On the Sony (original) output everything is fine (beautiful should I add) but on the modified output I get a high freq buzzing at very low level (I have to put my ear against the speaker to hear it). I am suspecting some form of instability because the modified output is stripped of any filters and have two separate trafos - one for the new clock board and one for the discrete output stage.

All in all - I do not think that the problem is with the amp in this case.
 
decky said:
IThe Ci is a filter capacitor to serve as a high pass filter together with R3 (680 ohm in Brian's layout) adjusted to very low frequency - 47uF and 680ohm leads to f=1/(2PI*R3*Ci)=5Hz - therefore 4.7Hz will filter below 0.5Hz so might be not as efficient. The LM3886 data sheet aims at 7.2Hz with Ri and Ci combination of 1kohm and 22uF respectively.


Wrong. 4,7 uF would filter ten times higher, i.e. 50Hz but that is just in theory, you will not be able to hear the difference.
Both tantalum and electrolytic capacitors have ugly non-linearities, so it is worth fitting an MKP. Again, this amount of DC is not a problem, but rather the noise which you mention, coming only from the CD player, and it might be that you are getting ground loops.

Yes, the snow is indeed melting but it freezes in the evening so we have free ice-skating and traumatology tours!
 

decky

Member
2006-09-12 6:10 am
Oouups I got my maths wrong :shutup: - You are right it is 50Hz .

However, I am still confused abot the measuring method - should I short the input when measuring the DC offset?:confused:

Rgerading the CD player noise - yes it does sound like higher harmonics hum noise. I will check the grounding in the player and probably put that MKP cap.

Thanks for your help
 

Sangram

Moderator
Paid Member
2002-09-25 11:01 am
India
Just to throw in my experience, I've found that one can eliminate Ci and have close to zero offset (<2mV) with shorted input.

Basically it requires very low value of Rf and Ri, I used 4.7K/220 without Ci on a 4766 and got output offset of 0.7/1mv (left/right).

With regular configuration of 22K/1K and a 47uF Ci the offset rose to about 10/13mV.

Eliminating Ci and not changing the resistor value gave me offset value of 60/85mV.

There was an interesting discussion with peranders on this issue, where I learnt that output offset voltage is a function of input offset currents. Also there is an interesting variation built by a forum member without Ci and 2.2K/100ohm feedback resistors, where sound quality was claimed to be improved.

Following are what I tried:

1. LM4766 - 22K/1K/47uF
2. LM4766 - 22K/1K/no Ci
3. LM4766 - 4.7K/220/no Ci
4. LM4780 - 22k/1K/47uF
5. LM4780 - 22K/1k/22uF
6. LM4780 - 10K/470/no Ci
7. LM4780 - 22K/1k/no Ci and BPA 4 chips/chanel
7. LM3886 - 22K/1K/47uf (latest, still settling in till I get bored of it)

Because of tin ears or poor speakers, I cannot tell the difference between any of the above configurations (within the same chip) with a good Ci (I am using FC and MKP parallel combination). But with a mass-market Ci, I could hear a difference. Not sufficiently qualified to state if it was degradation or enhancement (sounded like the former to me). There are significant differences between chips though, but that would be a different discussion.

FWIW all amps ran off an identical supply - 22VAC (32VDC) and with identical load (same speaker was used for the years across which all these amps were built, run, tested and blown up). The 4780s had PCBs, the rest are P2P amps.
 

decky

Member
2006-09-12 6:10 am
OK just another report and some extra info.

I have been listening my amp for last couple of days for couple of hours a day. The sound is still very good. I repeated all DC offset measurements just a minute ago. Also, I built the amp so it has a set of input bypass RCAs. The idea behind these is to use them to run subwoofers (one at the moment), since I have only one output from the preamp. Results are:

- input shorted ; no sub connected ; no load on output

Left- 17mV Right - 16.8mV

- input shorted ; no sub ; speakers connected

almost the same as above

- preamp on the input; no sub; with speakers

L- 100mV R-75mV

- preamp connected; sub on Right ; with speakers

L-100mV R-30mV

-open input ; no sub - with speakers

L-125mV R-95mV

- open input ; sub on Right ; with speakers

L-125mV R-35mV

It is obvious that shorted input makes a lot of difference and that the sub influences the DC offset due to some coupling with the crossover impedance load (not sure?).

Any comments and/or explanation more then welcome.
 
Your numbers suggest the preamp is not providing a low DC impedance at the input of the amp. That's not uncommon if the preamp uses output caps, or if it's a passive unit with just a pot to control the level.

Much confusion can be saved in testing if you follow the rule, "test it the way you use it". If the amp is going to live with a high output impedance preamp, that's the way you should test it. Not that you shouldn't test it in various other configurations to see what it will do.

There are offset voltages and offset currents. Find an op-amp book (like Jung's) and read it a few times- it takes a while to grasp. The resistor values around the amp will affect the offset, as will the driving impedance.

IMO, for headphones I don't like to see more than 10mV of offset. For larger speakers, 50mV is a good maximum, though a bit arbitrary.

If you wanted, you could add a pot to trim out the offset completely, though this seems to have fallen out of favor.
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.