The new DIY preamp designed during pandemic

[pma]

The new DIY preamp designed during pandemic

I needed a preamp with several single ended and balanced inputs, so it was a challenge during this strange pandemic times to design and make a new audio toy that would satisfy my needs. You cans see the result sitting between the Vincent CD player and DacMagicPlus D/A converter, in the next photo.

Design goals

• 3 single ended, 2 balanced inputs
• 1 balanced output usable as a single ended output as well
• very wide bandwidth that would not degrade hi-res sources
• excellent immunity to input HF and EMI
• considerable reduction of HF interference content arriving from digital signal sources
• high output current and ability to drive 600 ohm load impedance
• improvement of S/N and distortion compared to digital volume control

Circuit design

To fulfill the goals stated here above, the following circuit was designed

The inputs are switched by high quality relays. Single ended inputs go, after input RC filtration, directly to input buffer (IC9, IC101). Balanced inputs go, after common mode and differential RC filtration, to instrumentation amplifiers (IC1 – IC4) and then to the input buffer. Volume control (P1, blue Alps 2x10k/log) is connected behind the input buffer and effectively reduces noise of the preceding integrated circuits at lower volume settings. This is important to get the best S/N at low volume settings.
IC7, IC8, IC102 and IC103 are the output stage opamps. All the opamps used have very high BW and slew rate, that is 125V/us and higher. The slowest parts are the instrumentation amplifiers at balanced inputs wit slew rate of 15V/us, which is, to my measures, quite low value. They are protected from very high frequency input signals by input common mode and differential RC filters.

I do emphasize the speed of the opamps for the reason that IME the influence of input EMI is just the reason of different sound of different opamps.

Photos of preamp construction

This is the photo of partially assembled main board under square wave test

The main board inside the box. The box was bought at https://modushop.biz/site/, together with panel drilling and engraving.

This is the completely assembled box with wiring. One can see 2 power supplies, one for the electronics and the second one for relays. This reduces interference coupling from relay coils to the signal path. There is a common ground point for these 2 power supplies.

This photo shows the completed preamp.

Technical parameters

3 single ende inputs, SE1 – SE3
2 balanced inputs, BAL1 and BAL2
1 balanced output, usable as single ended as well
Input impedance SE 100 kohm
Input impedance balanced 136 kohm
Output impedance SE 47 ohm
Output impedance balanced 94 ohm
Max. output current 67 mAp
Max. input voltage SE1-3, BAL1 9 Vrms
Max. input voltage BAL2 4.14 Vrms
Max. output voltage SE 9 Vrms
Max. output voltage balanced 18 Vrms
Gain SE1-3, BAL1 inputs 0 dB to SE output, +6 dB to balanced output
Gain BAL2 input +6.62 dB to SE output, +12.62 dB to balanced output
Frequency range SE inputs 1.6 Hz – 1.06 MHz / -3dB
Frequency range BAL inputs 2.3 Hz – 134 kHz / -3dB
Rise time SE inputs 330 ns
Rise time BAL inputs 2.6 us
S/N SE in to BAL out 116 dBA / 4 Vrms at full volume
129 dBA / full output
118 dBA / 4 Vrms volume at 11:00 am
S/N BAL in to BAL out 106 dBA / 4 Vrms at full volume
119 dBA / full output
118 dBA / 4 Vrms volume at 11:00 am
THD I am not able to measure it as my system limit is about 0.0005%

Some measurements

Step response for 20Vp-p, SE input. Blue = input, Red = output.

100 kHz clipping, SE input.

Speaking about distortion, my system limit is about THD = 0.0005% and the preamp measurement is a copy of the soundcard loopback result. So I can only say that the distortion is lower. Next plot shows that there is no usual rise of distortion at higher frequencies, due to fast opamps used.

However, I do not consider distortion measurement as important if it is lower than 0.01%. There is no evidence that distortion below 0.1% would be audible, if it contains only low order harmonics. So it makes no sense to discuss about distortions that are less than 0.001% and to consider them as a measure of quality.

 

[Speedskater]

It appears like the XLR pin#1's are attached to the audio circuit common, rather than to the chassis at each connector.

 

[boXem]

Nice!

 

[muslhead]

Excellent
Are all outputs active at the same time?

 

[LTig]

Wow, looks absolutely professional. Great work.

Would you mind sharing the part numbers of the opamps and INAs?

 

[gene_stl]

Kit?? Circuit board?

 

[pma]

Excellent
Are all outputs active at the same time?

There is only one output connector, the balanced one, but it is possible to use a XLR - RCA adapter cable that takes the signal from OUT+ and GND. However, you can only use either balanced or SE output, not both at once.

 

[pma]

It appears like the XLR pin#1's are attached to the audio circuit common, rather than to the chassis at each connector.

This will be investigated yet. In theory or if only balanced signals are used exclusively, pin 1 is best on chassis. However, if there is a mix of SE and balanced signals connected, this recommendation may not work better. I am stating that based on many systems built and measured.

 

[pma]

Wow, looks absolutely professional. Great work.

Would you mind sharing the part numbers of the opamps and INAs?

INA217, AD825, LM6171, AD844.

 

[Feanor]

Wow, neat. Wish I could design and also build something like that.

No remote control though, pity.

 

[restorer-john]

@pma Very nice build, Pavel. You have been busy!

Interesting you have C57 and C102 connected to the wiper of the Alps pot which means the frequency response will vary depending on volume position, due to the source impedance changing the RC characteristics. What is the frequency response deviation at very low volume settings compared to wide open?

IC9 and IC101 can then also put a DC offset onto the pot track which may cause crackling in time.

Why wouldn't you place C57/102 on pin 6?

There is also potential for DC on the main preamplifier output which would affect DC coupled power amplifiers. You have no switch on/off muting that I can see. Is there a position on the selector that mutes all inputs? Can you show us a switch on/off capture of the output?

 

[pma]

Why wouldn't you place C57/102 on pin 6?

There is also potential for DC on the main preamplifier output which would affect DC coupled power amplifiers. You have no switch on/off muting that I can see. Is there a position on the selector that mutes all inputs? Can you show us a switch on/off capture of the output?

C57/102 is not on pin 6 because the Alps has 10k impedance and I would need 10uF cap. Now I use 1uF foil cap followed by 100k resistor, which makes 1.59 Hz/-3dB. I would never use and electrolyte as a coupling cap.

You are right about the possible DC output offset. In case of my LM6171 choice it is true (because of input bias current), but I do not care as my power amps all have input coupling foil capacitors. And the preamp is only for me. To prevent this, FET input opamps would completely fix this, namely my favorite OPA627, however it is not fast enough and I want my preamp to be as fast as possible and I have good reasons for it. The initial DC thump is decent. Moreover, the preamp and power amp are turned on simultaneously and the power amp has both delayed speaker SSR relays and output DC protection. So the start up and turn off are both absolutely silent.

My philosophy is to evaluate an audio chain as a whole. Measurements of isolated components like DACs and preamps or power amps is fine, but does not tell much. Only if you connect signal source, preamp and power amp together you get the result. And it should have best S/N and absolutely no hum and noise audible even with your ear at the speaker. There are components with fantastic measurements when measured isolated and have issues when interconnected in the whole chain. Especially 2 components in class I with signal ground connected to PE are a usual source of problems. I could name a DAC which has top results in Amir's measurements and makes big troubles when connected with class I amplifier. SINAD is easily degraded of 50 - 60dB. That's one of the reasons why I do not take 1kHz SINAD as a serious measure of audio quality.

 

[pma]

@restorer-john I made that turn-on transient measurement for you, John

And the HF output spectrum up to 10MHz as well.

 

[restorer-john]

C57/102 is not on pin 6 because the Alps has 10k impedance and I would need 10uF cap. Now I use 1uF foil cap followed by 100k resistor, which makes 1.59 Hz/-3dB. I would never use and electrolyte as a coupling cap.

Not knowing your Alps pot value, I figured it had to be low. Your input relays also shunt all the unused input signals to ground via a resistor to cut crosstalk/noise. I like that.

It's interesting, using a unity gain buffer and driving a low value pot and a high input impedance next stage. And you have excellent noise numbers too. It shows there are many ways to skin a cat.

I really like the instrumentation amp front end for the balanced inputs. All in all, a classy build for your purpose. I have to agree with you regarding whole system vs individual components. You're right, a properly earthed product vs unearthed Class ii metal chassis products- things often go astray. I'm old school, I like anything metal earthed and proper grounding practices followed.

So is that an +80mV offset after 300mS on the output?

 

[AnalogSteph]

I really like the instrumentation amp front end for the balanced inputs.

Moi, not so much. INAs have a lowest possible gain of unity by design, at which point the impedance seen by the inputs tends to get at least moderately high. INA217, being designed for microphone inputs, also features a very hot input pair with copious current noise. Not the very most ideal combination, and definitely a waste of some performance.

IMHO: If you need a balanced line receiver, use one. Like one of these. You can get them with gains of 0 dB, (+/)-3 dB and (+/)-6 dB, allowing for a unity gain bal in / bal out. And if -106 dBu of noise output is not good enough, you can always go discrete with a buffered low-impedance BAL/SE converter, Douglas Self style (using hand-matched resistors, not a major issue for the DIYer).

S/N BAL in to BAL out 106 dBA / 4 Vrms at full volume
119 dBA / full output
118 dBA / 4 Vrms volume at 11:00 am

BAL1 or BAL2?

BTW, if I am not mistaken, input noise of your balanced inputs comes out to about -98 dBu if that's for BAL1. That's actually slightly better than spec for the INA217, from which I would expect about -96 dBu at unity gain. I guess it's got a tiny bit of gain to make up for losses later on?

My philosophy is to evaluate an audio chain as a whole. Measurements of isolated components like DACs and preamps or power amps is fine, but does not tell much. Only if you connect signal source, preamp and power amp together you get the result. And it should have best S/N and absolutely no hum and noise audible even with your ear at the speaker.

ACK. As the proverb says, the proof of the pudding is in the eating.

There are components with fantastic measurements when measured isolated and have issues when interconnected in the whole chain. Especially 2 components in class I with signal ground connected to PE are a usual source of problems. I could name a DAC which has top results in Amir's measurements and makes big troubles when connected with class I amplifier. SINAD is easily degraded of 50 - 60dB.

Welcome to the world of the computer audiophile, where exactly this is a common problem. I'm still counting "why does <active studio monitor> hum/buzz/hiss/whatever if connected to my computer with unbalanced cabling?" as a FAQ.

IMO, IEC Class I amplifiers with (conventional, not Bruno Putzeys style) unbalanced inputs might just as well be listed in the dictionary under "hopeless". Big name hi-fi manufacturers had that figured out by the 1970s. In those days, people didn't have computers but they did have choonas, and said radio-receiving fish was commonly connected via coax to an external antenna which in turn was earthed (a typical legal requirement for lightning protection). That would have meant instant ground loop. Grounded sources are a fact of life, and amplifying electronics better deal with them.

Yes, Class II is harder for the DIYer to make safe (with some constructions you're quite happy that they're Class I...), and mains transformers with shield windings are not what one would call super easy to find. Tough luck. Hint: If you have too much mains leakage over the transformer, you can "cheat" with a Y capacitor to PE. Even a modest 10 nF would knock down resulting voltage considerably (seeing that even a rather big xfmr rarely makes it past 1 nF of leakage C), and even though I would expect few issues in terms of ground loops with such a small value, you can add ~1 kOhm in series to keep those at bay if needed.

If, however, you were talking about a DAC with a balanced output and an amplifier with balanced input, then that would be interesting. It would indicate a severe Pin 1 Problem. That would be a good one to know about if it's actually the DAC's fault, in which case its designer would deserve a printed copy of AES48-2005 to be thrown at them.

Life could be so much easier if audio equipment just stuck with AES48 compliant balanced interconnects. *sigh* That would still leave a whole bunch of different levels to be sorted out, but it would be a start.

That's one of the reasons why I do not take 1kHz SINAD as a serious measure of audio quality.

Ehhhh.... NACK.

Instead let's say SINAD is a necessary criterion, but not a sufficient one. You do also want to know about dynamic range and multitone distortion at the very least. AES48 compliance actually isn't an easy one to test for, although it has been done (Bill Whitlock et.al., I think).

 

[pma]

BAL1 or BAL2?

BTW, if I am not mistaken, input noise of your balanced inputs comes out to about -98 dBu if that's for BAL1. That's actually slightly better than spec for the INA217, from which I would expect about -96 dBu at unity gain. I guess it's got a tiny bit of gain to make up for losses later on?



Welcome to the world of the computer audiophile, where exactly this is a common problem. I'm still counting "why does <active studio monitor> hum/buzz/hiss/whatever if connected to my computer with unbalanced cabling?" as a FAQ.

IMO, IEC Class I amplifiers with (conventional, not Bruno Putzeys style) unbalanced inputs might just as well be listed in the dictionary under "hopeless". Big name hi-fi manufacturers had that figured out by the 1970s. In those days, people didn't have computers but they did have choonas, and said radio-receiving fish was commonly connected via coax to an external antenna which in turn was earthed (a typical legal requirement for lightning protection). That would have meant instant ground loop. Grounded sources are a fact of life, and amplifying electronics better deal with them.

Yes, Class II is harder for the DIYer to make safe (with some constructions you're quite happy that they're Class I...), and mains transformers with shield windings are not what one would call super easy to find. Tough luck. Hint: If you have too much mains leakage over the transformer, you can "cheat" with a Y capacitor to PE. Even a modest 10 nF would knock down resulting voltage considerably (seeing that even a rather big xfmr rarely makes it past 1 nF of leakage C), and even though I would expect few issues in terms of ground loops with such a small value, you can add ~1 kOhm in series to keep those at bay if needed.

If, however, you were talking about a DAC with a balanced output and an amplifier with balanced input, then that would be interesting. It would indicate a severe Pin 1 Problem. That would be a good one to know about if it's actually the DAC's fault, in which case its designer would deserve a printed copy of AES48-2005 to be thrown at them.

Life could be so much easier if audio equipment just stuck with AES48 compliant balanced interconnects. *sigh* That would still leave a whole bunch of different levels to be sorted out, but it would be a start.

Ehhhh.... NACK.

Instead let's say SINAD is a necessary criterion, but not a sufficient one. You do also want to know about dynamic range and multitone distortion at the very least. AES48 compliance actually isn't an easy one to test for, although it has been done (Bill Whitlock et.al., I think).

You are raising some good points and I can see you know what you are speaking about and I appreciate it.

1) "BAL1 or BAL2?"
I measure the S/N according to usual procedure, that means noise voltage at the output with volume pot at full gain position and input shorted. Under such conditions, the result for BAL1 and BAL2 is equal as output noise of the INA217 remains almost same regardless gain it has. This is basically due to resistor net noise inside the structure. However, real practically usable S/N is better with BAL2, as it has +6.62dB gain contrary to 0dB of BAL1. 2nd important thing is that the volume pot is behind the INA. So the permanent output noise of the INA is decreased as volume pot position is set to "normal" listening. BAL2 remains better as the volume may be set lower for the same output.
In fact the preamp remains absolutely quiet when connected to power amp with gain of some +27dB and speakers with some 86dB/2.83V/1m sensitivity.

2) "dBV or dBu"?
The results are really in dBV, calibrated before measurements. However, it is dBV(A), A-weighted, that I emphasize in the specs. The A-weighted result is always a bit better, compared to unweighted one. I use A-weight as a rule, to prevent confusion of noise numbers measured at different sampling frequencies and different BW. Same applies to SINAD and THD+N, which is in fact the same paremeter.

3) "class I issues"
you are describing correct points. I have designed a preamp with only "true" balanced inputs and outputs, and yes, pin 1 connected to chassis is the best way. However, consumer audio brings to us all varieties of nonsense, which in fact defines the resulting parameters, after the whole system is interconnected. This is uneasy to fix for non-professionals.
BTW, I hate RCA connectors. For my own purposes, I use the Neutrik XLR even for single ended connects. The worst properties of RCA are that it disconnects ground before the live and mechanical tolerances of connectors of different producers, that do not enable a reliable contact. I do not care that XLR is bigger, I do not care about miniature products.

Just for fun, attached is the measurement of very low level CD signal. Standard 44.1kHz/16bit, signal level is -100dBFS (yes it is below CD dithered noise over audio band), of course dithered. It is played by Vincent CD-S3 player, then it goes via coax to DacMagicPlus D/A, from D/A balanced output to preamp BAL1 input and then measured at the balanced output of the preamp. One can see a nice -100dBFS line rising from noise floor measured with spectral density defined as can be seen in the plot. The CD 16bit dithered noise prevails the system noise, as can be seen from spectrum above 20kHz, where the CD noise disappears.

 

[jasonhanjk]

Balanced input are noisier than unbalanced. Consider unbalanced signal while using XLR?

 

[pma]

Balanced input are noisier than unbalanced. Consider unbalanced signal while using XLR?

No, because for standard balanced sources with rated output of 4Vrms the S/N is more than high enough for practical listening purposes.

 

[Speedskater]

Balanced input are noisier than unbalanced. Consider unbalanced signal while using XLR?

You are considering a balanced stage rather than a complete interconnect system.

 

[jasonhanjk]

You meant the op amp configuration between non-inverting vs differential?