Moving from 9038Q2M to a 9028Pro board

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Markw4,

Isn't there a way to cure that LT3042 LF noise? A different or larger cap perhaps?

There are a couple of things. Best way is probably given in one of the LT3042 or LT3045 data sheets. See diagram below. Main problem I have found with using LT6655 is that it requires a physically big 10uf film cap for best performance, but I think Mikett found a prebuild LT6655 module that looks like it would work. Mike?

The usual way people try is less effective, and can be hard to retrofit. It involves increasing Cset in the LT3042 data sheet to the maximum value of 22uf. However, that will increase start up time of the regulator to a few or several minutes. Fixing that requires using the fast start up capability of LT3042 also described in the data sheet using the power good (PG) terminal.
 

Attachments

  • LT3045_with LTC6655reference.jpg
    LT3045_with LTC6655reference.jpg
    63.1 KB · Views: 583
Last edited:
During the modification of my 9038Q2M dac as on this thread ES9038Q2M Board - diyAudio

I accidentally killed my cheap DAC. This provided me an opportunity to decide whether to start all over again with the above DAC or start over with another kit but having perhaps less requirement for fabrication.

I looked at various DAC boards on Ebay and elected to not get something too costly because I could very well destroy it again and if I did. I would simply purchase something and be done with it.

Via the above thread I realized that substantial fabrication would be required and I did not wish to go to making boards again. With the cheap Es9038q2m mentioned, one needs to fabricate an op amp PS for AVCC as well as fabricate a new IV section for the DAC as the single op amp circuit on the board is simply not good despite putting on better op amps.

Then I saw a DAC board that looked interesting.
It has a proper IV converter section. I chose to stick with the 9028pro because the PS requirements were a lot less than a 9038Pro. Next, I wanted to get an DAC where I could easily insert improved PS sections that without too much trouble. I also desired a display.
The board I ended up with I think is a reasonable cost. It is a 9028pro board and the same design appears to be also used for the 9038pro. It can be found on numerous offerings on Ebay and AliExpress and appears to be a popular board. It uses 3 terminal regulators for the op amp supply...which easily allows me to insert improved PSs like a Sulzer which I already had or put in even a Super regulator for which I might still have some parts if I wanted to build one. It has one compromise. Left and Right AVCC is fed from a common regulator., has a display and a remote.

The project is not finished but this would be a project as it progresses I will report.

Warning, this thread will have a minimum of measurements as I don't have the equipment. So any reports of any improvement as I mod needs to be looked at as not fact but opinion. I state that up front.

After having received my DAC board, in stock form it had 5532/34 op amps. These provided a strong sense of midbass with a thick sound. High frequencies were slight muted. Changing these to LME49720 and LME49710 op amps removed the heavy midbass and provided more extended or more transparent sound. I had noted somewhat similar characteristics on the cheaper DAC but it was much more noticeable now.

The cheaper 9038q2M DAC before I killed it, had been modified. A lowZ low noise Sulzer Power supply was feeding the op amp and an LT3042 was feeding the DAC section. The LT3042 had 15uF of film caps for Cset. The DAC section had been modified by putting in 470uF OSCON sepc caps for decoupling.
The op amp IV section on the cheap DAC was in voltage mode with a single op amp section.

Now the stock 9028pro with the switch to the LME497xx op amps easily easily beat out the 9038q2m modded as above in the following areas. Lower octaves were more solid and defined. Separate voices were more distinct and transparent. This possibly points to the aspect that an ESS based DAC must have a proper IV section and the DAC must be operating in current mode for it to be good.

So as time progresses I will press forward by installing improved power supplies and possibly upgrading clocks and possibly upgrading on board components to better match ESS recommended circuits.

First I will install an improved power supply for the op amps.

Hi, could help me how to connect my 2 transformers to this DAC? Your first image helped a little, but I'm in doubt about what + and - in the DAC connectors. I say it for the 6v (2 connectors with 2 pins each ) and 15v (1 connector (right) with 3 pins) input.

I could not locate this info nowhere.
 
Mikett,

Just curious: have you compared or tried the LT3042 regulator versus the Super Regulator?

Which LT3042 board or kit are you using?

Markw4,

Isn't there a way to cure that LT3042 LF noise? A different or larger cap perhaps?

I have been busy and did not notice this.
A super Reg will not be easy to implement for the AVCC because you will need to find an op amp that can work withing the output voltage and that is 3.3V.
You might be able to, but it is going to be difficult.

I have not had the time to put together the LTC6655 LME49600/LME49710 variant as yet but I soon will. Whether that sounds better or not, I don;t know. However each mod I do, I try and keep the system in a state that I can easily backtrack.

For me there's no rush on this as in the current state it sounds nice to me and there is no urge to push ahead but I will anyways just to see.
 
You should be able to backtrack a bit and find the inputs to the op amps for the dual op amps. Take the op amps out and tap the signal from the socket if desired. You can take a normal op amp socket, push a couple pins out and solder the pins to the two leads on the transformer. Then stick the pins into the op amp socket. Do you have a schematic for the board from your vendor? This should allow a reversible mod and see if you like the result.
 
There are a couple of things. Best way is probably given in one of the LT3042 or LT3045 data sheets. See diagram below. Main problem I have found with using LT6655 is that it requires a physically big 10uf film cap for best performance, but I think Mikett found a prebuild LT6655 module that looks like it would work. Mike?

The usual way people try is less effective, and can be hard to retrofit. It involves increasing Cset in the LT3042 data sheet to the maximum value of 22uf. However, that will increase start up time of the regulator to a few or several minutes. Fixing that requires using the fast start up capability of LT3042 also described in the data sheet using the power good (PG) terminal.

The use of LTC6655 with LT3045 is, to some degre, a conundrum. There are
no measurements of the total (wideband in nV/rt Hz) noise performance on
either devices data sheet.

What we can probably assume, is that the VLF noise limitations of LT3045
are largely a result of ref current noise // set resistor noise as opposed to
the internal opamp itself.

As such a simpler way to get better 1/f noise is just use a much bigger, say x
10 = 220uF set pin capacitor and a pull up circuit on start up.

There are a few fairly simple ways to do this.

Are there any other reasons than slow start up why the set cap is limited to
22uF?

T
 
Are there any other reasons than slow start up why the set cap is limited to
22uF?

Just what it says in the data sheet. More than 22uf reduces regulation loop bandwidth, maybe a concern if one wants fast response time. Probably better to do as they recommend as use LT6655 as the reference.

In addition, the performance graphs of LT304x don't show what happens with noise below 10Hz. However, we happen to care about that because LF AVCC noise will intermodulate with the dac analog outputs to increase distortion.

Might be a good idea to use one LT6655 as a reference for multiple LT304x 3.3v regulators: one for the clock, one for AVCC (if using a PRO chip), maybe one for VCCA. Probably would work very well, have to try it and see though.

Also, probably a good idea if trying such a thing to bear in mind that LT304x regulators are precision high gain, high frequency devices which can cause problems if layouts and wiring are not well implemented. Some interesting commentary here: https://www.diyaudio.com/forums/power-supplies/311453-tried-chinese-lt3042-board-4.html#post5561404
 
Just what it says in the data sheet. More than 22uf reduces regulation loop bandwidth, maybe a concern if one wants fast response time. Probably better to do as they recommend as use LT6655 as the reference.

This doesn't make sense on many levels.

Provided the cap on set pin has suitably low ESR, it won't affect the HF loop
BW. The obvious solution is to parallel another smaller cap, if that is a concern.

In addition, the performance graphs of LT304x don't show what happens with noise below 10Hz.

That is exactly what I am saying. If there is a benefit in the sub 10Hz
region in using LTC6655, then the LT3045 MUST not be the bottle neck or
it would dominate.

The circuit you have posted above from data sheet merely over rides the self noise of the 3045's CCS and the set resistor. It's internal opamp noise can
not be reduced by this arrangement.

However, we happen to care about that because LF AVCC noise will intermodulate with the dac analog outputs to increase distortion.

Might be a good idea to use one LT6655 as a reference for multiple LT304x 3.3v regulators: one for the clock, one for AVCC (if using a PRO chip), maybe one for VCCA. Probably would work very well, have to try it and see though.

Not a bad idea, however I would be using more aggressive filtering on it's
OP by at least a factor of 2.

Also, probably a good idea if trying such a thing to bear in mind that LT304x regulators are precision high gain, high frequency devices which can cause problems if layouts and wiring are not well implemented. Some interesting commentary here: https://www.diyaudio.com/forums/power-supplies/311453-tried-chinese-lt3042-board-4.html#post5561404

[/QUOTE]

What he is saying is very true, so in many cases of non local use, good idea
to have OP damped. Data sheets should give more info on this.
 
...Just curious: have you compared or tried the LT3042 regulator versus the Super Regulator?

...A super Reg will not be easy to implement for the AVCC because you will need to find an op amp that can work withing the output voltage and that is 3.3V...

If you ran a +V super reg* off the the +-15V supplies; link out D2, connect Pos Vout to Pos Vout sense, connect Pos return sense to 0V; replace the Vref with a NiCd**, set the gain to 2.75 (x1.2V=3.3V).

The output can only source current, it cant sink - unless you add a shunt (think single ended output stage), is this what you need?

*The Superregs | Linear Audio NL
**https://tf.nist.gov/general/pdf/1133.pdf
 
Hi idir,
Those of us who have been here awhile all know each other and we know about each other's threads. Mikett, who started this thread also posts in the other 'ES9038Q2M board' thread. No need for you to tell us all about each other's threads. Thank you though, for your interest.

Mark
 
Hi idir,
Those of us who have been here awhile all know each other and we know about each other's threads. Mikett, who started this thread also posts in the other 'ES9038Q2M board' thread. No need for you to tell us all about each other's threads. Thank you though, for your interest.

Mark

Mark You are welcome. I did not inform you about your friend's thread, but about mine, What I would like to achive is to have a simple to do thread.

For example
1. AVCC: method, pictures, comment. Done
2. Output stage, schematics, explanation, pictures done.
I will be biulding my own piece of both boards with the mobile dac and the PRO, so might be able to document that. With your help guys if possible.
 
Understood, I meant that Mikett subscribes to both threads so he would likely see your posts either way.

Also, it is likely a mistake to treat all mods as though they are all a good idea. Some of them may not do much good, or are a distraction from making more effective progress. For that reason, it may confuse people more than help to put every possible mod all in one thread.
 
Last edited:
I have a few questions:
1. What would be the benefits of using 7805 instead of 7808 as pre-regulator. Less heat for LT3402 ?

2. Cset capacitor LT3402 are we talking about the caps i red circles? is MURATA X7R apropriate? As per datasheet lowest noise is achieved using capacitance between 4.7 uF and 10 uF. why use 22 uF? In the graph there is a sliught increase from 10 uf upwards.
noise2_edited.jpg
 
I have a few questions:
1. What would be the benefits of using 7805 instead of 7808 as pre-regulator. Less heat for LT3402 ?

2. Cset capacitor LT3402 are we talking about the caps i red circles? is MURATA X7R apropriate? As per datasheet lowest noise is achieved using capacitance between 4.7 uF and 10 uF. why use 22 uF? In the graph there is a sliught increase from 10 uf upwards.
View attachment 726201

Cset is the cap connected to the set pin. It is the cap not circled in red. 22uf is the maximum recommended value to minimize low frequency noise, but it also slows down start up time. Therefore, the fast start up option should probably be used if Cset is increased to 22uf.

5V vs 8v would reduce power dissipated in secondary regulators. LDOs don't usually need much extra voltage, although a little more like 8v might be good in some cases. Haven't really tried to see if it helps anything.
 
Cset is the cap connected to the set pin. It is the cap not circled in red. 22uf is the maximum recommended value to minimize low frequency noise, but it also slows down start up time. Therefore, the fast start up option should probably be used if Cset is increased to 22uf.

5V vs 8v would reduce power dissipated in secondary regulators. LDOs don't usually need much extra voltage, although a little more like 8v might be good in some cases. Haven't really tried to see if it helps anything.

Just started back up again.
5V IME is not enough for a current boosted LT3042 like the one I used. Someone else has also found this as well. I would say 8V is the minimum allowing about a little under 5V overhead. I'm going with 9V.

I have further upgraded the AVCC reg I am currently using with an AD797 on the Sulzer and it is a definitely a step up in detail and lower distortion. Image depth improved as well.

I have also received my AD8031s and was able to solder this small op amp onto the Super Reg board directly. So I willl slowly build up my Super Reg 3.3V AVCC using an LTC6655 as a voltage reference and AD8031 as the op amp. The AD8031 is fast slews well. I expect it to be better than the Sulzer.

Over the holidays I also was able to test an AD797 op amp driving the comparable on a PCM1794. I found that the power supply feeding AVCC op amp is also important. With a Super Reg, the op amp is powered with the output of the Super Reg which it is controlling itself. So I've found no need to provide a high quality pre reg there.

That AVCC I feel is so critical on DAC with current outputs both on the BB and ESS dacs I have sampled. Can't be stressed enough. My PCM1794 is using an AD797 with 5V reference derived from a 7805 whose noise is way higher than the prior LTseries LDO it replaced but it sound so much better. Be careful with the LDO regs super low noise regs. Forget the specs, listen.
 
Hi All,

Last yearr i purchased the q2m board and did several modifications on it.
(Opa replacement. Cap replacement, avcc feed with own opa based solution, iv stage, full ps separation with lp5907, salas reg for analog part, also posted in the other thread).
Now i purchased a 9038pro board. As i have seen several modification should be done on it (when arrives)
My question;
Could someone send me a link or pdf wiht 9038pro datasheet and application notes?
Could someone summarize the best resulted modifaction experience with the pro board?

Thanks

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